[Admin Console] : Add cluster-aware /admin/metrics SSE endpoint (Prometheus-backed)

[Ferryx349]

Add GET /admin/metrics as an authenticated SSE stream for the admin console.

Metrics are sourced from Prometheus (OTEL -> Collector -> Prometheus).
Scope

• Authenticated admin-only SSE endpoint (will reuse Phase 1 admin session/auth)
• Periodic snapshots built from PromQL queries against existing OTEL metrics, e.g.:
  • Events/sec: rate(nostream_events_accepted_total) (& rejected if needed)
  • Accepted / rejected totals (counter values or rates)
  • Active WebSocket connections: nostream_websocket_connections
  • CPU / memory per worker (from nostream_process_* gauges; aggregate or sum by worker_type / worker_index as needed)
• Include lightweight DB and Redis health in each snapshot
• Cluster-wide view comes from Prometheus aggregation (worker labels on OTEL resource attrs), not primary -> worker IPC

Depends on

• OTEL bootstrap (feat(observability): add OpenTelemetry metrics bootstrap with Prometheus sidecar #653)
• Handler instrumentation ([Admin Console] : Add instrument event and websocket handlers for metrics. #654)
• Admin auth foundation (Phase 1 / feat: add admin backend foundation (login, session, health) #641)
• Prometheus reachable from the relay process (compose env or PROMETHEUS_URL)