Description of problem:

When uninstalling the config-policy-controller addon, a pod called `config-policy-controller-uninstall` gets deployed which sets the `policy.open-cluster-management.io/uninstalling` annotation to `"true"`. Because of a fix in ACM-6941 due to a misbehaving OADP version update, the governance-policy-addon-controller started setting the uninstalling annotation to `"false"` when deploying. It was supposed to set this to `"true"` when the addon is being deleted/uninstalled, but the addon framework used by the governance-policy-addon-controller does not propagate the change to the ManifestWork when it is being uninstalled.

This can lead to the uninstall stalling because the ManifestWork from the governance-policy-addon-controller and the config-policy-controller-uninstall pod are fighting each other.

This was noticed in the Service Delivery environment.

Version-Release number of selected component (if applicable):

2.8.1

2.9.0

How reproducible:

Reproducible at times.

Steps to Reproduce:

1. Install the config-policy-controller addon
2. Have at least one ConfigurationPolicy with pruneObjectBehavior set to a value other than None.
3. Delete the config-policy-controller addon
4. If the config-policy-controller-uninstall pod runs and restarts every 5 minutes without completing successfully, the issue has been recreated. You can check that the ManifestWork overwrote the uninstalling annotation.

Actual results:

Uninstalls can stall.

Expected results:

Uninstalls don't stall.

Additional info:

This probably also affects the governance-policy-framework addon.