Reconcile machine deletion hooks

Task:

As outlined in the quorum protection proposal the etcd-operator should add a deletion hook to the Machine CustomResource corresponding to a learner member node, and remove the hook when a machine’s node has been removed from the etcd cluster membership.

While currently part of the ClusterMemberController, this should ideally be added as a separate controller within the etcd-operator.

Acceptance Criteria:

1. Ensure that all voting members in an existing cluster have pre-drain deletion hooks on their corresponding machines
2. Add a new machine to the cluster and ensure the deletion hook is added to the machine (after its member is added as a learner)

Once promotion and scale-down capabilities have been added, hook removal can also be tested:

1. Delete a learner member’s machine, and check that the deletion hook for that machine is then removed (once the learner member has been scaled down)
2. For a cluster at the desired control-plane size (e.g 3 voting members), delete a voting member’s machine and ensure the Machine deletion hook remains and prevents the Machine from being deleted

https://github.com/openshift/cluster-kube-apiserver-operator/pull/1309