Container HEALTHCHECK and last-run status for orchestration

[rfsbraz]

Description

The Dockerfile ends at CMD ["python", "-m", "app.deleterr"] - no HEALTHCHECK. In scheduler mode there is no way for Docker/Unraid/k8s to detect a dead scheduler thread, a crashed-but-running container, or a string of failed runs. The run summary only goes to logs.

Proposed Behavior

• Persist a small status file after every run attempt: timestamp, success/failure, libraries processed/failed, items deleted/tagged
• Add a HEALTHCHECK script that fails when the process is unhealthy or the last successful run is older than N x the configured interval
• The same status file gives dashboards/scripts something machine-readable to consume (a /metrics-style endpoint can build on it later)

Implementation Notes

• Ties into Crashed scheduled runs are reported as successful, with no notification #293 (crashed runs currently leave no trace) and Scheduler silently skips missed runs (no misfire grace, no coalescing) #292 (missed runs)
• Keep it dependency-free: a JSON file plus a tiny check script is enough for v1