Revive retained zero-replica Deployments when a drained version becomes draining again

Worker Controller can scale a drained version’s Kubernetes Deployment to 0 after scaledownDelay, then delete it later after deleteDelay.

We should verify what happens if, during that retention window, the version becomes Draining again because someone moves workflows onto it with a Versioning Override.

Expected behaviour: while the Deployment still exists, the controller should be able to scale it back up so workers can poll for that version again and resume/complete work. There is some existing logic that may already handle this for rollback/target-version cases, but we should add a regression test to confirm the drained-to-zero Deployment can be revived before deleteDelay expires. If the test fails, we may need a small adjustment to the scaling logic.

Start by locating the existing scaling logic for rollback and target-version cases, then trace how a retained Deployment moves from zero replicas back to Draining. Add a regression test covering a drained-to-zero Deployment becoming Draining again before deleteDelay expires; done means the Deployment scales up so workers can resume or complete work.