feat(DM01-6276): add keepAlive option to GKE cluster service

[liuwei08]

Summary

Adds a keepAlive option to the GKECluster service spec, allowing clusters to persist after the InfraBox job ends. This enables teams to use long-running GKE clusters for debugging without needing manual provisioning.

Closes DM01-6276 — Vsystem team needs a long-running GKE 1.37 cluster to debug InfraBox job compatibility issues.

Changes

• types.go: Add KeepAlive bool field to GKEClusterSpec
• handler.go createCluster: Append --labels infrabox-keep-alive=true to the gcloud create command when keepAlive: true
• handler.go deleteGKECluster: Skip cluster deletion and only remove the CRD when keepAlive: true
• handler.go getOutdatedClusters: Exclude clusters with infrabox-keep-alive=true label from GC loop

Usage

"services": [{
  "apiVersion": "gcp.service.infrabox.net/v1alpha1",
  "kind": "GKECluster",
  "metadata": { "name": "my-cluster" },
  "spec": {
    "zone": "us-east1-b",
    "machineType": "n1-standard-1",
    "numNodes": 1,
    "diskSize": 100,
    "clusterVersion": "1.37",
    "keepAlive": true
  }
}]

After the job ends, the GKE cluster continues running in GCP. The user is responsible for manually deleting it via gcloud container clusters delete.

Test plan

• Create a job with keepAlive: true and verify cluster is not deleted after job finishes
• Create a job with keepAlive: false (default) and verify cluster is deleted as before
• Verify GC loop does not clean up clusters tagged with infrabox-keep-alive=true
• Verify clusterVersion field works together with keepAlive