Allow dismissing known CI failures from 'needs attention' list

[costajohnt]

Problem

When a PR has a CI failure caused by upstream infrastructure (not by the PR's code), the daily check keeps flagging it as "needs attention" on every /oss run — even after the user has already investigated, determined it's not actionable, and filed an upstream issue.

Example: Homebrew/brew#21594 has a syntax check failing because cache-homebrew-prefix overwrites the PR's source code with stale cached code (tracked in Homebrew/actions#796). There's nothing to do on our end, but /oss keeps listing it as a CI failure requiring attention.

Expected Behavior

After investigating a CI failure, the user should be able to acknowledge/dismiss it so it no longer appears in the actionable issues list. It could still show in the dashboard with an "acknowledged" or "snoozed" label.

Possible Approaches

1. Acknowledge command — oss-autopilot acknowledge <pr-url> --reason "upstream issue" that marks the issue as known
2. Snooze with expiry — dismiss for N days, then re-surface in case the upstream fix landed
3. Classify as infrastructure — if the CI check is classified as infrastructure rather than actionable, skip it automatically

Snooze with expiry (option 2) feels like the best fit — the upstream issue might get fixed, and you'd want the PR to re-appear when it does.