fix: retry failed latest only

[mdiapenabc]

Description

Fixes the csa-retry-failed cron replaying the full historical job_runs backlog (e.g. 83 failed rows / ~2+ hours on dev) instead of only retrying jobs that still need attention.

Root cause: getFailedJobs() returned every top-level failed cron job with no deduplication, and RETRY_FAILED itself could be retried recursively.

Changes:

• Exclude RETRY_FAILED from the retry queue to prevent recursion
• Retry only the latest failure per job_type (not the full backlog)
• Skip cron failures superseded by a later success (completedAt <= lastSuccess)
• Align hasStuckOrFailedJobs() with the same actionable set so the hourly cron exits early when there is nothing to do

End-user SEND_CRA_FILE failures are unchanged: latest failure per type is still retried regardless of later successes.

Fixes # (issue)

How Has This Been Tested?

Ran npm test for:

• jobs.service.spec.ts
• job-runner.service.spec.ts
• retry-failed.handler.spec.ts

Manual verification on dev (post-deploy):

1. Confirm next csa-retry-failed run logs a small retry count (≈ number of job types with current failures, not 80+)
2. Confirm latest INGEST_DATA failure still retries while ICM is down
3. Confirm cron exits early when only stale FAILED rows remain in job_runs

Follow UP

• Delete - Historical stale FAILED rows remain in job_runs (harmless; they are just no longer retried)