cluster-maxunavailable should not abstain from installing its hook on machines of a paused cluster

There is no real reason for the misc-controller-suite cluster-maxunavailable controller to do nothing on a paused cluster: adding the pre-drain hook is something that is perfectly safe to do on a paused cluster without risking to change its actual state.

And "do nothing on a paused cluster" is actually preventing us to properly use this controller in release-1.4 (see !5478 (comment 2754243418)).