feat(health): /healthz + /readyz + gRPC health for container deploys

[randomizedcoder]

Summary

Adds first-class liveness / readiness so xtcp2 is easy to run under Docker and Kubernetes.

Stacked on #61 (base = feat/listener-ip-ttl) to avoid conflicts in initPromHandler / grpc_server.go. GitHub will retarget this to main once #61 merges — review/merge #61 first.

What

• pkg/health — a process-wide readiness flag + HTTP handlers, served on the existing metrics listener:
  • /healthz — liveness: 200 as soon as the server is up (k8s livenessProbe).
  • /readyz — readiness: 200 only after the daemon passes its DestinationReady + netlinkers-up gate (i.e. it's actually polling), 503 before that and during shutdown (k8s readinessProbe/startupProbe).
• gRPC — registers the standard grpc.health.v1 service on the gRPC port, starting NOT_SERVING and flipping to SERVING on the same readiness condition (native k8s gRPC probes).
• The poller flips readiness (x.setReady) the moment it passes the DestinationReady gate, and back on return; setReady drives the HTTP flag and the gRPC status together.
• Bumps goVendorHash for the new google.golang.org/grpc/health import.

No new flags — the endpoints are always on, on the existing listeners.

Verification

go test ./pkg/health/          # /healthz always 200; /readyz 503->200 on SetReady
nix build .#xtcp2-s3parquet
nix build .#test-cmd-xtcp2
nix build .#test-go-flavor-s3parquet

Readiness reflects real daemon state (destination initialised + netlinkers started), so an orchestrator holds traffic/rollout until xtcp2 is genuinely collecting — not merely that the process is up.

🤖 Generated with Claude Code