feat: add rate-limited retry and observability for per-project GC partition setup

[zachsmith1]

Per-project GC partitions could silently fail to initialize during bulk startup when client-side throttling caused discovery or syncMonitors calls to fail. There was no retry mechanism, leaving affected projects with incomplete monitors permanently.

• Refactor AddProject to register-after-verify, preventing zombie builders
• Add rate-limited work queue to projectprovider with exponential backoff
  and configurable retry limits for project onboarding
• Add resyncNeeded flag to force GC resync when new partitions are added
• Add Prometheus metrics for partition health, project add outcomes,
  retry counts, queue depth, and add duration
• Expose project-provider tuning via CLI flags (workers, max-retries,
  rate-limit, rate-burst)

[joggrbot]

📝 Documentation Analysis

All docs are up to date! 🎉

---

✅ Latest commit analyzed: b556ad5 | Powered by Joggr