fix: Add Helm hooks for schema Job to fix deployment ordering

[jiechenz]

What was changed

Added Helm hooks (pre-install, pre-upgrade) to the schema Job and shims ConfigMap to enforce deployment ordering. The schema Job now completes before server pods are created.

Why?

When using Elasticsearch for visibility, a race condition exists between the schema Job and server Deployments. If a server pod writes to Elasticsearch before the schema Job creates the index with correct mappings, Elasticsearch auto-creates the index with wrong field types (e.g., WorkflowId as text instead of keyword). The schema Job then fails with mapper [WorkflowId] cannot be changed from type [text] to [keyword].

Helm hooks eliminate this race by ensuring the schema Job runs and completes before any server Deployments are created.

Checklist

1. Closes [Bug] Error in temporal-schema pod creating mappings for Elasticsearch #882

2. How was this tested:

• helm unittest
• helm template verified hook annotations present
• Deployed on kind cluster with PostgreSQL + Elasticsearch:
  • Verified schema Job completes before server pods start
• Deployed via ArgoCD on kind cluster:
  • verified argoCD correctly mapped Helm hooks to PreSync wave

3. Any docs updates needed?

Updated README with a new section "Schema Setup and Deployment Ordering"

[robholland]

nit: We should soften this if the hooks can also help in some ArgoCD cases, which the README suggests is the case. Maybe call out the specific for ArgoCD if it's succinct enough, or refer to the README if not.

[jiechenz]

Removed ArgoCD from the comments.