Add Asynq queue metrics (depth, active, lag, retries, DLQ)

[tayebmokni]

Summary

Expose Asynq queue metrics from §5.3 (queue depth, active, job duration, retries, failures, DLQ size, processing lag). These metrics drive the worker HPA (#81) and the §12 SLO alerts on processing lag and DLQ growth.

Design reference

• docs/10-observability.md §5.3 (Background jobs section)
• docs/12-jobs-cron.md §11.3 (Metrics)

Acceptance criteria

• gonext_asynq_queue_depth{queue}, gonext_asynq_active_jobs{queue} gauges
• gonext_asynq_job_duration_seconds{task_type} histogram
• gonext_asynq_job_retries_total{task_type} counter
• gonext_asynq_job_failed_total{task_type, kind} counter (kind: error/panic/timeout)
• gonext_asynq_dlq_size gauge
• gonext_asynq_processing_lag_seconds{queue} gauge — age of oldest pending job
• gonext_jobs_enqueued_total{type,queue}, gonext_jobs_idempotency_skips_total{type}, gonext_jobs_unique_conflicts_total{type} counters
• task_type cardinality bounded (~30) — enforced by registry (Public-form CSRF tokens (HMAC + anon-cookie binding) #176)
• Tests verify metrics emitted on enqueue, success, failure, retry

Dependencies

#150, #176

Complexity

M

[tayebmokni]

Closing — shipped in the final-wave PR (#487-#492). Auto-close only fires for first ref.