Port GridSelectorModal/CardSlot clusters onto DisplayPage

[WilfordGrimley]

Description

Wave 3 of issue #272's parity work (following wave 1's PR #395 and wave 2's
PR #403 — the final unparked tranche from wave 2's own "deferred to a future
wave" list): un-skips and ports 30 of the 55 remaining Playwright tests from
the classic /editor grid's coverage onto the unified DisplayPage at
/editor, stacked on proposal-h-editor-display-route-swap (PR #389) the
same way waves 1-2 were.

Ported clusters (30 tests)

1. GridSelectorModal.spec.ts (23 tests) + GridSelectorModalVariants.spec.ts
   (7 tests) — full port. GridSelectorModal.tsx's only surviving mount
   post-route-swap is CardbackToolbarButton/CommonCardback.tsx's
   project-wide cardback picker (testid cardback-grid-selector, title
   "Select Cardback") — per-slot picking now goes entirely through the
   rail's own Select Version section instead (a different component,
   already covered by SelectVersionSection.spec.ts). GridSelectorModal.tsx
   itself is completely generic (a bare imageIdentifiers array + onClick
   callback, doesn't care what the identifiers represent), so every
   grouping/filter/keyboard/mobile-viewport behavior these two files test is
   identical regardless of which caller's identifiers feed it — this is the
   full-fidelity instance both files port onto. New
   openDisplayCardbackGridSelector helper (test-utils.ts); new
   cardback-endpoint fixtures reusing the existing cardDocuments* sets
   (cardbacksThreeResults/cardbacksFourResults/cardbacksTwoSources/
   cardbacksWithCanonicalCards, mocks/handlers.ts).
2. CardSlot.spec.ts (15 of 25 tests) — delete/duplicate/change-query/
   right-click-context-menu/auto-select-first-result coverage, retargeted
   onto the sheet's own page-preview-slot + page-preview-slot-menu-cue
   (new openDisplaySlotMenu/openDisplaySlotContextMenu helpers — the
   sheet's visible "..." menu cue is the direct equivalent of the classic
   3-dot button, both open the identical card-slot-context-menu).
3. CardSlot.visual.spec.ts (2 of 6 tests) — the two grid-selector
   aria-snapshot tests, retargeted onto the cardback picker (see Add NotMPC ordering tab and PringlePrints listing #1).

Deviations from a clean port

• CardSlot.spec.ts: 10 of 25 tests dropped, each with its own comment in
  the file:
  • "switching to the next/previous image" + "wraps around" (3 tests) — the
    classic grid's inline ❯/❮ cycling arrows have no equivalent anywhere on
    the unified page; per-slot picking is browse-and-click only now (Select
    Version), with no "next/previous"/wrap-around concept to port onto.
  • "selecting an image in a CardSlot via the grid selector" — investigated
    and confirmed a genuine, previously-undocumented product gap, not
    dropped for convenience: this test's real payload is
    docs/features/card-dom-api.md's contract (data-card-* attributes +
    the mpc:card-selected event). Confirmed by grep against
    common/cardDom.ts's own callers: Card.tsx/CardSlot.tsx/
    CardDetailedViewModal.tsx all wire it; PagePreview.tsx (the unified
    page's own sheet-slot renderer) does not, at all — no data-card-*
    attributes, no event dispatch. Flagged in both docs/features/ card-dom-api.md and docs/features/grid-selector.md (edited in place)
    for an owner priority call, same as wave 1's CardImageStates.spec.ts
    finding.
  • Bulk multi-select (double-click-to-select-all + both shift-click tests +
    "most recently selected card is tracked") — 4 tests — no unified-page
    equivalent (issue /display parity inventory: verified remaining gaps vs /editor (switchover checklist) #272 item 6, still not built; SelectedImagesRibbon.spec.ts
    itself has been parked, not ported, every wave).
  • requested-printing badge "plain style"/"degraded style" (2 tests) —
    already covered verbatim by DisplayPage.spec.ts's own two badge tests
    against the identical shared RequestedPrintingBadge component in the
    rail header (the sheet slot itself never renders this badge —
    PagePreviewSlotContent carries no such field). Porting again would
    just duplicate coverage, same precedent as wave 2's
    DeckbuilderConfirmAffordance.spec.ts drop. The third badge test (the
    "absent" case) is NOT covered elsewhere and was ported.
• CardSlot.visual.spec.ts: 4 of 6 tests dropped — the per-slot inline
  candidate-strip snapshots (thumbnails + "N / M" counter + ❮/❯ arrows) have
  no sheet equivalent; same finding wave 1's import cluster already made for
  selectedImage/totalImages.
• CardSlot.visual.spec.ts's 2 ported snapshots are regex-tolerant, not
  exact-matched, on a handful of specific leaves (icon-only chevron-toggle
  buttons, the "Filters"/"Collapse All" buttons' own leading icon, "Remove
  None"/"Remove Source" chip buttons' trailing clear-icon, each
  StyledDropdownTreeSelect's placeholder-caret text node) — verified
  directly: a plain toMatchAriaSnapshot polling assertion reproducibly
  settles on a different accessible-name/text value for exactly these nodes
  than a single --update-snapshots capture does, deterministically and
  independent of worker count/explicit wait length (both ruled out by direct
  testing across 10+ runs) — a genuine pre-existing rendering race in this
  third-party icon-font/tree-select machinery (unchanged by this port), not
  a route-swap or port artifact. Every other leaf (headings, table rows,
  source names, result-tile names/images, DPI/file-size bounds) matched
  exactly and consistently across every run. Full reasoning in the file's
  own module comment.

New findings flagged for the owner (not ported, not silently dropped)

1. docs/features/card-dom-api.md's DOM API contract is unimplemented on
   /editor's sheet slots — see that doc's own "Known gap" entry
   (edited in place) for the full detail and a scoped fix description.
2. Per-slot next/prev image-cycling has no unified-page equivalent — see
   docs/features/grid-selector.md's own new "Known gaps" entry.

Neither is tracked against any numbered item in issue #272's own checklist —
both are new findings from this wave, not previously-known gaps.

Verification

• All 30 newly-ported tests green, run individually per-cluster together
  (npx playwright test tests/CardSlot.spec.ts tests/visual/CardSlot.visual.spec.ts tests/GridSelectorModal.spec.ts tests/GridSelectorModalVariants.spec.ts) —
  69/69 passed alongside the wave-1/2 marquee regression files
  (ImportXML.spec.ts, chunkErrorRecovery.spec.ts,
  DisplayLeftRailFidelity.spec.ts, SelectVersionSection.spec.ts).
• visual/CardSlot.visual.spec.ts's 2 ported tests specifically re-verified
  for stability: 6/6 green across 3 repeats × 4 parallel workers, after
  landing on the regex-tolerant fix described above (this took real
  investigation — the flake was 100% reproducible before that fix, not
  timing-sensitive; the reasoning is captured in the file's own module
  comment).
• Full suite (npx playwright test, excl. perf): 269 passed, 36 skipped
  (owner-parked clusters, untouched), 6 failed on the one full-suite run —
  all 6 re-run individually and passed: ImportCSV.spec.ts and
  CardDetailedViewModal.visual.spec.ts (generic import-text textbox
  timeouts under this run's heavy full-suite load, files this PR never
  touched), Navbar.spec.ts's Sign-out test and UnsavedWorkGuard.spec.ts's
  editor->/print test (both already documented as pre-existing/environment
  flakes in wave 1's own PR Port import/card-detail/export test clusters onto DisplayPage #395 description), visual/DynamicLogo.visual.spec.ts
  (documented dedicated-port-workaround artifact, also from wave 1's PR).
• npx tsc --noEmit clean.
• npx jest --silent — 568/568 passing (unchanged; no src/ files touched
  besides the new mocks/handlers.ts fixtures).
• npx prettier@2.7.1 --check clean on all changed files.

Environment note: local runs used a temporary, uncommitted
playwright.config.ts port override (3427) per docs/lessons.md's
worktree-port-collision entry, reverted before every commit (confirmed via
git diff showing no changes to that file in the final commit).

Open items for the owner

1. The two new findings above (DOM API gap, no per-slot image-cycling
   equivalent) — worth their own tracked issue(s), separate from /display parity inventory: verified remaining gaps vs /editor (switchover checklist) #272?
2. CardImageStates.spec.ts (wave 1's own flagged gap) and
   DeckbuilderConfirmAffordance.spec.ts (wave 2's own investigated-and-
   dropped file) remain untouched, as before.
3. The parked clusters (visual/CommonCardback.visual.spec.ts,
   SelectedImagesRibbon.spec.ts, ProjectEditorMobileScroll.spec.ts,
   PDFGenerator.spec.ts, PagePreview.spec.ts,
   PostExportContributionPrompt.spec.ts) remain untouched, skips intact,
   per the owner's own prior parking decision.

With this wave, the 55-test "deferred to a future wave" tranche wave 2's PR
#403 flagged is now: 30 ported, 14 dropped with justification (bulk
multi-select/duplicate-coverage/no-equivalent-UI), 1 dropped as a confirmed
product gap, 10 remaining in CardSlot.visual.spec.ts dropped as the same
class of "no sheet equivalent" as the wave-1 import-cluster precedent — no
tests silently left in limbo.

Checklist

• I have installed pre-commit and installed the hooks with
  pre-commit install before creating any commits.
• I have updated any related tests for code I modified or added new
  tests where appropriate. (This PR is the test update.)
• I have manually tested my changes as follows: see Verification above.
• I have updated any relevant documentation or created new
  documentation where appropriate (docs/features/card-dom-api.md,
  docs/features/grid-selector.md, both edited in place).

---

Stacked on proposal-h-editor-display-route-swap (PR #389, held open,
parity-blocked), same stacking as waves 1-2's PRs #395/#403.