feat(helm): opt-in PodDisruptionBudget and probes for CNPG plugin sidecars

[WentingWu666666]

Draft PR for review/discussion only. Part of the GA-readiness chart audit (#381).

What this PR does

Two small, additive, opt-in improvements for pod availability.

PodDisruptionBudget (M9)

Adds templates/11_pdb.yaml, gated by podDisruptionBudget.enabled (default: false).

Disabled by default on purpose: the operator currently runs at replicaCount: 1. A PDB with minAvailable: 1 (or maxUnavailable: 0) on a single-replica deployment blocks node drains forever that's worse than no PDB. Users running multi-replica with leader election should set podDisruptionBudget.enabled=true. When the operator chart gets multi-replica defaults in the future, this default can be flipped.

Plugin probes (M12)

The sidecar-injector and wal-replica deployments are gRPC servers on port 9090 with no probes today pods are marked Ready as soon as the container starts, regardless of whether the gRPC endpoint is actually serving.

Adds tcpSocket readiness + liveness probes on port 9090, gated by pluginProbes.enabled (default: true) with tunable initialDelaySeconds / periodSeconds / failureThreshold.

TCP socket probe is used because the plugins don't expose an HTTP health endpoint. It verifies the gRPC server has bound the port and is accepting connections better than nothing, not as strong as a real gRPC health check (deferred until the plugins implement grpc.health.v1).

Local verification

• helm lint clean.
• helm template renders PDB only when podDisruptionBudget.enabled=true.
• helm template renders TCP probes by default; omits them with pluginProbes.enabled=false.
• helm upgrade on kind with --set podDisruptionBudget.enabled=true creates the PDB; subsequent --set podDisruptionBudget.enabled=false removes it cleanly.
• Sidecar-injector pod becomes Ready (TCP probe passes against the actually-running gRPC server, confirming the probe target is correct).
• Pre-existing user-supplied values are preserved across helm upgrade (Helm default behavior; documented in PR for reviewer awareness).

Tracking

• Helm chart improvements for GA readiness #381 (GA chart audit, items M9 and M12)
• feat(helm): add pod security, resources, and scheduling knobs to chart #382 (in-flight: pod security/resources/scheduling)
• feat(helm): preflight check for cert-manager presence #383 (in-flight: cert-manager preflight)

[documentdb-triage-tool]

🤖 Auto-triaged by documentdb-triage-tool.

Applied: enhancement
Project fields suggested: Component manifests · Priority P2 · Effort M · Status In Progress
Confidence: 0.85 (mixed)

Reasoning

effort from diff stats (70+0 LOC, 4 files); LLM: Additive opt-in Helm chart improvements (PodDisruptionBudget and plugin probes) as part of GA-readiness audit, touching manifests/templates with no schema or cross-component changes.

If a label is wrong, remove it manually and ping @patty-chow so the rules can be tuned. The bot will not re-label items that already have component labels.