Replica: fix(webhooks): add scheduled task to unlock stale webhook request locks

[lucaforni]

Questa PR replica la PR originale: postalserver#3546

Autore originale: @cymulatereouven
Branch originale: fix/tidy-stale-webhook-request-locks
Repository originale: cymulatereouven/postal

---

Summary

• Problem: When a worker process crashes or is killed (common in Kubernetes with rolling updates, OOM kills, pod evictions), locks held on webhook_requests rows are never released. Unlike QueuedMessage which has TidyQueuedMessagesTask to clean up stale locks, WebhookRequest had no equivalent cleanup mechanism — causing webhook delivery to be permanently blocked.
• Fix: Adds a TidyWebhookRequestsTask scheduled task (runs hourly) that finds webhook requests locked for more than 1 hour and unlocks them so they can be retried.
• Adds a with_stale_lock scope on WebhookRequest (mirrors the existing pattern on QueuedMessage)

Changes

File	Change
app/models/webhook_request.rb	Added with_stale_lock scope (locks older than 1 hour)
app/scheduled_tasks/tidy_webhook_requests_task.rb	New scheduled task to unlock stale webhook requests
app/lib/worker/process.rb	Registered TidyWebhookRequestsTask in the worker TASKS array

Why unlock instead of destroy?

TidyQueuedMessagesTask destroys stale queued messages because they represent outbound email delivery attempts that are likely no longer relevant. Webhook requests, however, carry event notifications that downstream systems may still need — so we unlock them to allow retry rather than silently dropping them.

Test plan

• Verify WebhookRequest.with_stale_lock returns only records with locked_at older than 1 hour
• Verify TidyWebhookRequestsTask unlocks stale records (sets locked_by and locked_at to nil)
• Verify previously-stuck webhook requests are picked up again by ProcessWebhookRequestsJob after unlock
• Confirm no impact on actively-locked (non-stale) webhook requests

🤖 Generated with Claude Code