feat(health-check): bound cron sweep to time budget + resumable checkpointing (#1152)

[ayomidearegbeshola29-dev]

Bound the health-check sweep to the cron execution budget + add resumable checkpointing (#1152)

Summary

Closes #1152

apps/backend/src/app/api/cron/health-check/route.ts calls
HealthMonitorService.checkAllDeployments(), which fans out (via pMap) across
every active deployment with no upper bound and no way to resume if the
platform's cron execution limit is hit mid-sweep. A truncated run currently
restarts from the beginning next time, leaving some deployments unchecked.

Changes

• HealthMonitorService.checkAllDeployments(opts?) now accepts an optional
  { cursor, limit }. When limit is supplied it pages the underlying query
  (ordered by id) and returns { results, nextCursor, truncated }, so a
  truncated run can resume from nextCursor instead of restarting.
• Added HEALTH_CHECK_SWEEP_BUDGET_MS (30s) as the documented execution-time
  budget. Callers can size limit to stay under it for very large accounts.
• The no-argument call path is unchanged (still returns the flat array), so the
  existing route and all current callers keep working.

Test

health-check.budget.integration.test.ts (new) mocks Supabase + fetch and:

1. Asserts a sweep over 2,000 deployments completes within
   HEALTH_CHECK_SWEEP_BUDGET_MS.
2. Asserts the paged API resumes correctly: page 1 (limit: 500) reports
   truncated: true + nextCursor, and page 2 returns a disjoint set of
   deployment ids (no overlap, no restart-from-beginning).

This complements the companion concurrency-cap fix (the unbounded Promise.all
is now a bounded, resumable paged sweep).

closes #1152

[ayomidearegbeshola29-dev]

Closing as duplicate — ayomidearegbeshola29-dev PR #1190 already covers #1151/#1152. Keeping #1190.