feat(monitoring): liveness/readiness probes, SLO alert rules, and monitoring docs

[zakariyaufarida5-wq]

Closes #367

Summary

The backend already had most of the logging & monitoring stack — structured JSON logs via the pino logger (src/config/nest-pino-logger.ts) with request/response logging middleware, Prometheus metrics defined in src/config/metrics.ts and recorded by src/metrics/metrics.interceptor.ts, and a GET /metrics endpoint plus a combined GET /health. This PR closes the remaining gaps in the acceptance criteria.

Changes

• Liveness & readiness probes (src/app.controller.ts) — split the single /health into orchestrator-grade probes:
  • GET /health/live — liveness; performs no dependency checks, so a transient downstream outage doesn't cause a pod restart.
  • GET /health/ready — readiness; verifies critical dependencies, so the orchestrator routes traffic only when the service is actually ready.
    Both are @Public() and un-rate-limited (probes run frequently).
• Prometheus SLO/SLI alert rules (monitoring/prometheus/alerts.yml) over the real exported metrics: error-rate (>5%), p95 request latency (>1s), p95 DB latency (>500ms), in-flight saturation, event-loop lag, auth-failure spikes, and TargetDown.
• Grafana dashboard (monitoring/grafana/stellaiverse-backend-dashboard.json) — request rate, error ratio, latency percentiles, saturation, DB latency, auth failures.
• Documentation (docs/monitoring.md) — log field format, full metrics table, health-endpoint usage (with example K8s probe config), and the alerting thresholds/SLOs.

Acceptance criteria mapping

• Structured JSON logs (requestId/userId/timestamp/severity) — already present; documented in docs/monitoring.md.
• Metrics (latency, error rates, saturation) + Prometheus endpoint — already present; documented.
• Health-check endpoints for liveness and readiness — added here.
• Dashboards and example SLO/SLI alert rules — added here.
• Documentation for logging format, metrics list, and alerting thresholds — added here.

Verified locally against the repo's CI checks: npx tsc --noEmit and npm run build both pass; the alert rules and dashboard were validated as well-formed YAML/JSON.