fix(k8s): make node metrics listener configurable

[jiangqiao2]

Problem

The Kubernetes Node DaemonSet runs with hostNetwork: true, while the Edge diagnostics/metrics listener was hard-coded to :9101. If the target node already has a host-level ongrid-edge process or another exporter using that port, the Node Edge fails to start and enters CrashLoopBackOff.

Reproduced in production: the host-level Edge was already listening on :9101, preventing the Kubernetes Node Edge from starting.

Fixes #261

Changes

• Added the ONGRID_EDGE_METRICS_ADDR environment variable. The default remains :9101 for backward compatibility.
• Controller, standalone Edge, and Kubernetes data-plane diagnostics now use the same configurable address.
• Added the Helm value node.metricsAddr and injected it into the Node DaemonSet.
• Included the metrics address in the DaemonSet configuration checksum so changing the value triggers a rolling update.
• Added tests covering the default value, environment variable override, and checksum changes.

Verification

• git diff --check
• go test ./cmd/ongrid-edge
• Full make test-k8s-chart
• Kubernetes verification: after setting node.metricsAddr=:19101, the Node reached 1/1 Running; :19101 and the existing :9101 listener operated simultaneously.

The local environment did not have Go or Helm installed. Unit tests and chart validation were completed on the target server with Go 1.25 and Helm.

• I have read and agree to the project contribution guide. Guide: https://github.com/ongridio/ongrid/blob/main/CONTRIBUTING.md

Port Configuration

The Node DaemonSet uses hostNetwork: true and listens on :9101 by default. If another service is using that port on a node, specify an unused port:

--set-string node.metricsAddr=:19101

The value can also be set in values.yaml:

node:
  metricsAddr: ":19101"

The selected port must be available on the host of every Kubernetes node.