fix(arrs): stop queue cleanup worker when arrs is disabled at runtime

[javi11]

Summary

• Toggling arrs.enabled (or arrs.queue_cleanup_enabled) off via config previously did not stop the already-running queue cleanup worker. It kept ticking on its configured interval and logging Failed to cleanup Sonarr queue against the now-disabled Sonarr/Radarr instances.
• Root cause: the gate check lived only in Worker.Start (internal/arrs/worker/worker.go); there was no OnConfigChange handler for the ARR subsystem, unlike the health subsystem which already does this.
• Fix mirrors the existing pattern in internal/health/controller.go:
  • New Service.RegisterConfigChangeHandler subscribes to config changes and starts/stops the worker when arrs.enabled or arrs.queue_cleanup_enabled flips.
  • New exported helper worker.IsQueueCleanupEnabled(cfg) centralizes the predicate.
  • Defensive early-return at the top of Worker.CleanupQueue so any in-flight tick after a disable no-ops cleanly.
  • Wired the handler in cmd/altmount/cmd/serve.go right after StartWorker, alongside the other OnConfigChange registrations.

Test plan

• Start with arrs.enabled: true and a reachable Sonarr instance — confirm ARR queue cleanup worker started appears.
• Set arrs.enabled: false (config.yaml or UI). Expect ARR worker disabled via config change, stopping followed by ARR queue cleanup worker stopped, and no further Failed to cleanup Sonarr queue warnings on subsequent ticks.
• Toggle back to true. Expect ARR worker enabled via config change, starting and the worker resumes.
• Verify the queue_cleanup_enabled: false sub-toggle (while arrs.enabled stays true) produces the same stop/start behavior.
• go build ./... and go vet ./... pass (verified locally).