Add real multi-repo GitHub replays

[2002yy]

What changed

• add a production-path recorder that stores compact GitHub review-context replay evidence without comment bodies, full source, or tokens
• check in real provider recordings for Flask #3709, Vite #144, and Gson #705 with independently reviewed immutable gold labels
• require provider replay source metadata to match manifest repository, PR, base SHA, and head SHA
• report the first cross-language real baseline in the project status document

Why

The curated seed cases were useful for deterministic mechanics but could not establish real mapping quality. These recordings expose historical-line loss, cross-fork identity, ambiguous mapping, removed review targets, provider latency, and partial status without presenting curated examples as representative evidence.

Baseline

• corpus: 4 repositories / 5 cases / 3 real provider replays
• real replay status: 3 partial, 30 provider requests, about 192.3s mean live recording latency
• aggregate review-symbol mapping: precision 0.25 / recall 0.1667 / F1 0.20
• CI association remains unproven on real failed-CI cases and is explicitly not claimed

Validation

• replay/recording/evaluation: 12 passed
• GitHub suite: 97 passed
• full pytest: 786 passed
• frontend: 49 files / 172 tests and production build passed
• Ruff: passed
• targeted mypy with skipped imports: passed
• package structure and secret scan: passed (1024 files)

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.