fix: support revision-scoped HA pod anti-affinity

[mary-ppv]

Description

Added optional revision-scoped pod anti-affinity to helm_lib_pod_anti_affinity_for_ha.

Callers can enable it using:

{{ include "helm_lib_pod_anti_affinity_for_ha" (list . (dict "app" "test") (dict "revisionScoped" true)) }}

When enabled, the helper adds:

matchLabelKeys:
- pod-template-hash

Existing calls remain unchanged because revision scoping is disabled by default.

Why do we need it?

During a Deployment rolling update, required pod anti-affinity based only on common application labels can prevent a pod from the new ReplicaSet from being scheduled alongside a pod from the old ReplicaSet.

Scoping the rule by pod-template-hash keeps replicas of the same revision distributed across different nodes, while allowing pods from different revisions to temporarily run on the same node. This prevents rollouts from becoming blocked during transitions such as HA to non-HA.

The behavior is opt-in to avoid changing scheduling semantics for existing users of the helper.

Testing

Added unit tests verifying that:

• existing calls do not render matchLabelKeys;
• calls with revisionScoped: true render pod-template-hash;
• pod anti-affinity is not rendered when HA is disabled.

All unit tests passed:

Test Suites: 83 passed
Tests: 358 passed

The generated Helm library documentation was updated.

Version

Bumped the chart patch version from 1.72.13 to 1.72.14.