feat: add healthcheck latency metrics and response fields

[paulobernardoaf]

What

Adds Prometheus histogram metrics for Redis, MongoDB, and total healthcheck duration, extends the /healthcheck JSON response with the same latency data, and shares ping helpers between the HTTP handler and the gRPC health monitor. The gRPC binary now exposes /metrics so those observations are scrapeable.

Why

Operators need visibility into dependency ping latency to detect slow or failing Redis/MongoDB connections before they impact clients. Recording metrics only for dependencies that were actually checked avoids polluting histograms with false zero-second MongoDB samples when Redis fails early in the gRPC monitor.

Test plan

• go test ./pkg/health/... ./pkg/metric/... ./pkg/websocket/ -run 'TestHealthStatus|TestRecordLatencies|TestObserveHealthcheck'
• Hit GET /healthcheck and confirm redis_latency_seconds, mongo_db_latency_seconds, and total_latency_seconds in the JSON body
• Scrape GET /metrics and confirm healthcheck_duration_seconds histogram with dependency label (redis, mongodb, total)
• Verify gRPC pod exposes /metrics and records health monitor observations

Made with Cursor