operator: does not clean up stale RetinaEndpoints across restarts

**Describe the bug**

`RetinaEndpoint` objects can leak across missed events in the operator, called out in https://github.com/microsoft/retina/blob/0a8044757e75e4104a0ca4c6ea25821854505320/pkg/controllers/operator/retinaendpoint/retinaendpoint_controller.go#L47.

**To Reproduce**

Stop operator pod, delete some pods, restart operator pod. Notice that the `RetinaEndpoints` for the deleted pods are not cleaned up.

**Expected behavior**

Any `RetinaEndpoint` objects no longer exist for deleted pods.

**Additional context**

Fixed by https://github.com/microsoft/retina/pull/2707.

Start with pkg/controllers/operator/retinaendpoint/retinaendpoint_controller.go at the referenced line and review the restart scenario: stop the operator, delete pods, and restart it. Compare the behavior with linked pull request #2707. Done means RetinaEndpoint objects for deleted pods are removed after missed events and an operator restart.

Written by the indexing model from the issue text.