fix(memory-ingest): --include-unattributed is inert — apply the fallback to page.git_remote (#2353)

[noron12234]

Fixes #2353.

Root cause

buildTranscriptPage applies the _unattributed fallback in two of the three
places it computes it, and the one it skips is the one the guard reads:

const remote = resolveGitRemote(session.cwd);   // "" when the remote can't be resolved
const slug_repo = repoSlug(remote);             // → "_unattributed"
…
`git_remote: ${remote || "_unattributed"}`,     // :729 frontmatter — defaulted
…
git_remote: remote,                             // :750 page record — RAW

if (!args.includeUnattributed && page.git_remote === "_unattributed") {   // :1173

page.git_remote is "", so the comparison is never true, the branch is dead, and
--include-unattributed has no effect in either direction. The earlier !session.cwd
guard doesn't cover it — these sessions have a cwd, it just no longer resolves to a
remote (deleted directories, ephemeral worktrees).

buildArtifactPage already stores the defaulted value (git_remote: slug_repo, :787),
so this is the transcript builder catching up.

Fix

One line, plus a comment noting the guard depends on it:

git_remote: remote || "_unattributed",

Taking the fix at the source rather than loosening the guard to !page.git_remote || … === "_unattributed" keeps the field consistent with the frontmatter and the slug,
which is what buildArtifactPage already does.

Verification

Reproduced with a real session JSONL whose cwd is a directory that isn't a git repo,
run through --bulk --no-write:

	written	skipped (unattrib)
before	1	0
after	0	1
after, --include-unattributed	1	0
after, cwd is a repo with a remote	1	0

Three regression tests added to test/gstack-memory-ingest.test.ts, in the existing
runScript / writeClaudeCodeSession style. Confirmed they actually bite — reverting
just the one-line change makes the first one fail:

(fail) unattributed skip > skips a session whose cwd is not a git repo
Expected to contain: "written:               0"
Received: "… written:               1 … skipped (unattrib):    0 …"

Suites run clean with the fix in place:

bun test test/gstack-memory-ingest.test.ts                  26 pass  0 fail
bun test test/memory-ingest-no-put_page.test.ts \
         test/memory-ingest-timeout.test.ts \
         test/gstack-memory-helpers.test.ts                 34 pass  0 fail

Note on scope

The issue also suggests recovering attribution from the Claude Code project-directory
encoding rather than discarding it. That's a behaviour change with its own design
questions, so it's left out — this PR only makes the existing flag do what it says.

[trunk-io]

Merging to main in this repository is managed by Trunk.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here