Report Pipeline Finding

The deployer's watchdog (ops/railway-ci/deployer/watchdog.py, release-path unit U6, design section 3.7) reports a release-pipeline condition it noticed on its 2-minute tick -- a dead fixer, a red main, staging or production falling behind, a stalled CI queue, a stale image build, ci-service drift, a ready pull request sitting unmerged, a stranded branch, PROMOTE_ALLOW_RED_CI used.

Deliberately the SAME channel Rooster's own findings already use (app.modules.rooster.supervision.record_finding, POST /api/rooster/supervision/report), not a second one: the System page and Frederick's Slack DMs already read that one card/DM path, and a release-pipeline finding is exactly the same shape -- an observed condition, deduped by reference so a still-red main is one open card, not one per 2-minute tick, and cleared automatically the next time supervision.record_resolution sees the reference reported healthy.

Gated on system:pipeline_findings:write (migration 202609301620, same two-part pattern #2515 used for the GitHub App token mint), not on rooster:supervision:write -- the deployer's principal is not Rooster's own pipeline principal and should not inherit whatever else that grant happens to cover.