feat(apisix): add configurable livenessProbe support

[feillu-a]

What this PR does

Adds an optional livenessProbe configuration to the APISIX chart.

Currently, only a readinessProbe (TCP on :9080) is configured. Without a
livenessProbe, a frozen APISIX process (no crash, but unresponsive) will
cause the pod to stay in Not Ready state indefinitely without ever being
restarted by Kubernetes.

Changes

• values.yaml: add livenessProbe block with enabled: false by default
• templates/deployment.yaml: inject the probe when enabled

How to test

# values.yaml
livenessProbe:
  enabled: true

The pod should now be restarted automatically if APISIX stops responding on :9080.

Closes #951

[martin-mat]

+1. Today the gateway only has a readiness probe, so a pod that stops serving is taken out of the Service's endpoints but never restarted. With one or two replicas that quietly leaves no gateway at all until someone deletes the pod by hand. An opt-in liveness probe fixes that without changing anything for existing installs. Once #1007 exposes the status endpoint, an httpGet on it would be the natural next step, since it also catches a hung worker whose port still accepts connections.