fix(usage): report incomplete transcript scans

[shivamhwp]

Usage summaries could omit unreadable transcripts while claiming complete coverage. Scans now keep readable records and report partial for entry or file failures, failed for an unreadable root, and missing for an absent root. Diagnostics contain bounded counts rather than raw filesystem errors.

Failed reads retain their previous parse cache for recovery, and incomplete directory walks cannot evict unseen cache entries. The shared client merge prefers a complete copy of the same source over a partial one and excludes failed sources from ownership.

Fixes #5798. Stacked on #11298. Codex cumulative-counter validation in #5800 remains separate; this change reports filesystem coverage failures.

Validation: 39 focused reader, service and merge tests pass, plus server/shared typechecks and scoped lint. Six service/merge regression scenarios fail on the original code. A separate Node replay with real Linux permission failures retained a readable file beside an inaccessible directory, reported an inaccessible root as failed, and recovered the original usage after permissions were restored. Directory rotation, file stat errors, retrying unchanged bytes and cache preservation are covered. Windows and macOS filesystem behavior remains unverified.

Internal refinement remains pending after the final authorized Fable retry failed. No merge requested.

Implemented and tested with GPT-6 through Codex in T3 Code.

[macroscopeapp]

Approvability

Verdict: Approved at 89f0222

Macroscope's review found this PR approvable — This is a focused usage-scanning bug fix that reports incomplete filesystem reads without discarding readable or cached transcript data. Normal successful scans remain unchanged, and targeted tests cover retry, partial walks, root failures, and duplicate-source selection.

^{You can add or adjust custom eligibility rules. Learn more.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	13.6 KiB	—	15.1 KiB	✅
Codex	Thread snapshot wire	—	7.0 KiB	—	7.3 KiB	✅
Codex	Live turn WebSocket wire	—	6.6 KiB	—	7.8 KiB	✅
Codex	Live turn WebSocket decoded	—	57.1 KiB	—	66.4 KiB	✅
Codex	Live turn messages	—	10	—	21	✅
Claude	Total thread wire	—	13.6 KiB	—	15.1 KiB	✅
Claude	Thread snapshot wire	—	7.1 KiB	—	7.3 KiB	✅
Claude	Live turn WebSocket wire	—	6.6 KiB	—	7.8 KiB	✅
Claude	Live turn WebSocket decoded	—	57.9 KiB	—	66.4 KiB	✅
Claude	Live turn messages	—	10	—	21	✅

Baseline: unavailable · PR result: 89f0222 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.9 KiB
• Claude decoded thread snapshot: 114.6 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.