feat: add retention policies and cleanup jobs for auth_nonces, DLQ, processed_events, agent_logs (#161)

[Nanle-code]

Closes #161

What changed

Added scheduled data retention jobs to prevent unbounded table growth across four tables, with Prometheus metrics and configurable retention periods via environment variables.

New files

• src/jobs/dataRetention.ts — cleanup functions for all four tables + scheduler
• prisma/migrations/20260618000000_add_retention_indexes/migration.sql — indexes to support efficient cleanup queries

Modified files

• src/config/env.ts — added config.retention block with four env-configurable retention settings
• src/utils/metrics.ts — added retention_deletes_total{table} and retention_last_run_timestamp_seconds{table}
• src/index.ts — registers scheduleDataRetention() after server starts; clears handle on graceful shutdown
• .env.example — documents all new retention env vars with defaults

Retention defaults (all overridable via env)

Table	Env var	Default
processed_events	RETENTION_PROCESSED_EVENTS_DAYS	90 days
dead_letter_events (RESOLVED only)	RETENTION_DEAD_LETTER_EVENTS_DAYS	30 days
agent_logs	RETENTION_AGENT_LOGS_DAYS	60 days
auth_nonces	(expires naturally via expiresAt)	immediate
Run interval	RETENTION_INTERVAL_MS	24h

Notes

• DLQ cleanup only removes RESOLVED records — PENDING and RETRIED are left intact
• Jobs run once on startup to clear any backlog, then on the configured interval
• All jobs follow the existing sessionCleanup pattern: try/catch, structured logging, metrics recording