Probe configured external dependencies (Redis, etc.)

Return per-dependency status for monitoring + post-deploy verification.

Currently probes Redis (rate-limiter backend). The endpoint is always 200 so dashboards can read individual statuses without losing the full picture to a single failed dep — mirror the boot-log behaviour where Redis errors fail-open rather than crash the app.

Response shape (additive — new deps appended over time): { "redis": { "configured": bool, # is REDIS_URL set? "reachable": bool | null, # null when not configured "latency_ms": float | null, "error": str | null, # categorical; never raw exception text } }