docs: fold fire-sequence measurement into pipeline-fidelity-gate.md

[WilfordGrimley]

Summary

• Folds the 2026-07-24 DB-verified per-pass measurement of the Pipeline-fidelity gate: dpi=250 parity replay + knowledge-inventory sweep (task #151) #154 fire sequence into docs/pipeline-fidelity-gate.md (the canonical status page), editing §2/§6/§9/§14 in place rather than appending a dated section.
• Corrects the live resolved-printing count from a provisional 4 back to 3 (a second consensus_recompute closer flipped one card resolved→unresolved after the doc's original snapshot).
• Adds the per-pass table (run_ids, wall times, rows written) for the five post-fire corrective/completion passes that extend true completion to 2026-07-24: lexicon-gate retraction, marker reparse, artist-credit fill, calculator re-pass, and the second consensus-recompute closer.
• Adds the pass-9 ledger gap as an honest, docs-sourced note (49,206 tag transitions have no PilotRunLedger row — structural, pre-ledger-convention gap, re-confirmed by exhaustive DB scan, not closeable).
• Adds the artist-consensus finding: all 218,345 cards remain artist_vote_status=unresolved because single-machine votes sit below the owner-ratified vote-weight resolution threshold (links to reference/vote-weight-matrix.md rather than restating it).
• Adds operational notes: the scryfall-cache-lost-on-rebuild incident (issue scryfall_cache is lost on every container rebuild #402) and border-attribute-tag seeding (7 tags created via seed_attribute_tags, DB-verified).
• Updates open items: Bug-A full re-scan (still deferred, tracked); moderation package sizing is now data-ready off the DB-verified review-queue size (134,370 cards).
• Updates the stale cross-references in docs/theory.md §7d (resolved count, artist finding) and the one-line index summaries in docs/README.md, docs/MANIFEST.md, and CLAUDE.md so nothing restates gate status inconsistently with the canonical page.

Context: issue #154 (pipeline-fidelity gate). Full canonical status lives at docs/pipeline-fidelity-gate.md per this repo's own convention — this PR keeps that single-source-of-truth property intact.

Test plan

• python3 .github/scripts/docs_lint.py --strict — clean
• python3 .github/scripts/tests/test_docs_lint.py — 26/26 pass
• python3 .github/scripts/tests/test_publish_wiki_link_rewrite.py — 4/4 pass (not a wiki-published page, but ran for parity anyway)
• python3 .github/scripts/tests/test_publish_readme.py + regenerate-and-diff against committed readme.md — no drift
• Every new/changed figure in docs/pipeline-fidelity-gate.md §14 cross-checked against a live, read-only DB query (docker exec mpcautofill_django python manage.py shell) against production Postgres, 2026-07-24 — including the border-tag-seeding claim (Tag ids 24–30) and the scryfall-cache issue (scryfall_cache is lost on every container rebuild #402, confirmed open via gh issue view)
• pre-commit hooks (prettier) ran clean on the final commit

Deviations

• None from the task instructions. One addition beyond the literal instruction list: verified the "border-tag seeding (7 created)" and "scryfall-cache loss incident (issue scryfall_cache is lost on every container rebuild #402)" operational notes independently against the live DB/GitHub API rather than taking them on faith, since neither was in the source measurement report — both checked out exactly as described.