feat: add queue depth alert webhook

[eclectic-coding]

Summary

• New alert_queue_thresholds config — a hash of queue_name => ready_job_count limits
• Fires a webhook when any configured queue's ready count meets or exceeds its threshold
• Uses the same alert_webhook_url endpoint(s) as the failure alert, with event: "queue_depth_threshold_exceeded" plus queue_name, depth, and threshold fields
• Cooldown tracked independently per queue — a persistently deep "critical" queue does not suppress alerts for "default"
• Triggered from DashboardController#index alongside the existing AlertWebhook

Implementation

• QueueDepthAlert service — self-contained, mirrors the structure of AlertWebhook; queries per-queue ready counts via a single grouped COUNT; per-queue @last_fired_at hash behind the same MUTEX pattern
• alert_queue_thresholds reader defaults to {} so unconfigured apps are unaffected
• 12 specs covering threshold boundary, per-queue independence, cooldown isolation, payload shape, and error handling

Test plan

• Set alert_queue_thresholds: { "critical" => 5 } — enqueue 5+ ready jobs to "critical" → webhook fires on next dashboard load
• Enqueue fewer than 5 → no webhook
• Enqueue to "default" (no threshold configured) → no webhook
• Two queues both over threshold → two webhooks fire
• Fire once, reload dashboard within cooldown → only one webhook total
• Cooldown expires for "critical" but not "default" → only "critical" fires again

🤖 Generated with Claude Code