[onedr0p]

Rollout PR #2 of the org metrics/health standard (primary-HTTP tier + the /healthz+/readyz pair).

What changed

• /healthz (liveness) + /readyz (readiness) on the main webhook port (8080). readyz aliases healthz today — the endpoint exists so probes and external configs never change if chaski ever gains a real serving condition. (This also retires the long-standing wart of /readyz appearing in the debug-log path map without being served.)
• Probes target the http port (liveness+startup → /healthz, readiness → /readyz). New chart test pins that metricsEnabled: false removes the metrics containerPort while probes still resolve.
• Metrics listener is metrics-only and truly optional — disabling removes the listener, container port, Service port, and ServiceMonitor. This supersedes the interim fix(server): always serve /healthz on the monitoring port; record panicked requests #38 shape (always-on monitoring listener serving health): health now rides the always-present main port, so there's no orphan 8081 when metrics are off.
• helm test hook now checks /readyz on the Service http port instead of metricsPort/healthz.
• CHASKI_METRICS_PORT stays a port int bound ":<port>" (already dual-stack safe).

Compatibility

• In-chart consumers (probes, ServiceMonitor) updated automatically.
• External probe/scrape configs pointing at :8081/healthz must move to the http port — hence feat!.

go test -race green, lint 0 issues, helm lint + helm unittest (20/20) green, docs/schema regenerated, helm template --set config.metricsEnabled=false renders no metrics port and http-targeted probes.

Prior PR in the rollout: home-operations/echo#28. Next: kromgo, konflate, then tuppr/litellm-operator/miroir, then drm-exporter.