fix(ci): fix deployment lockups and protect database PVCs

[DerekRoberts]

This PR resolves the ongoing failures in the Merge to Main deployment workflow and adds robust protections for database PVCs against accidental data loss.

Changes Included:

1. Fix main deploy lockup recovery: Removed the pull_request event gate from the pre-existing deployment cleanup step in .deploy.yml. This allows the Merge to Main workflow (triggered by workflow_run or workflow_dispatch) to detect and auto-rollback stuck releases.
2. Prevent mid-deploy cancellations: Set cancel-in-progress to false in the Merge to Main concurrency configuration. This prevents GitHub Actions from terminating deployments mid-run and leaving Helm releases in a locked pending state.
3. Fail-safe database PVC retention:
   • Modified values.yaml and pvc.yml templates to default to "helm.sh/resource-policy": keep for the database PVC.
   • Configured .deploy.yml to explicitly override database.pvc.keep=false only for ephemeral PR deployments (ensuring we clean up DEV resources).
   • Added a double-safe condition check and inline documentation to prevent accidental deletions on future workflows/dispatches.

[Copilot]

Pull request overview

This PR updates the reusable Helm deployment workflow to always attempt cleanup of stuck/pending Helm releases, not only on pull_request events, so non-PR deploy workflows (e.g., workflow_run / workflow_dispatch) can recover from “another operation is in progress”.

Changes:

• Removed the if: github.event_name == 'pull_request' gate from the “stop pre-existing deployments” cleanup step.
• Updated the inline comment to no longer imply the cleanup is PR-only.

[mishraomp]

very nice