process: ignored-test policy + bitrot audit (#826 followup)

[zackees]

Background

The #826 audit counted ~50+ #[ignore] annotations across the test suite. Most are legitimate (require attached hardware: a real ESP32, a Teensy, a USB CDC device) but some are stale — the underlying reason for the ignore was fixed and nobody removed the annotation. Without a tracking discipline, ignored tests rot until they no longer even compile.

Examples from #826

• crates/fbuild-build/tests/teensy30_acceptance.rs — #[ignore = "requires teensy hardware"] (legitimate; hardware required)
• crates/fbuild-build/tests/avr_build.rs — #[ignore] (reason unclear; needs investigation)
• crates/fbuild-packages/tests/disk_cache_schema_migration.rs — #[ignore] (reason unclear)

(Full list to be enumerated by the script proposed below.)

Why this is NOT a dylint

This is a process concern, not a code-shape concern. The right tool is a periodic audit, not a per-commit lint.

Proposal

1. Add ci/audit_ignored_tests.py that walks the workspace, finds every #[ignore] (with and without a reason string), and emits a CSV: file,test_name,reason,line.
2. Add a weekly GitHub Action (cron 0 0 * * MON) that runs the script and posts the diff vs. the previous week to a tracking issue (open one stable issue tracking: ignored-test inventory and update its body).
3. Define the policy:
   • Every #[ignore] must have a reason string (#[ignore = \"...\"]).
   • The reason must reference either a tracking issue (#NNN) OR a hardware requirement (e.g. requires teensy 3.0).
   • A test that's been ignored for 90+ days without a hardware reason gets escalated to a CodeRabbit review.

References

• dylint: candidate code-quality lints (gotcha sweep) #826 — original gotcha-sweep tracking issue
• PR for the extended-826 dylint set (this issue is the C2 follow-up from that PR)