Currently, rolling of ZooKeeper clusters is tightly coupled with StatefulSets. That is something what needs to be dealt with as part of the ZooKeeper removal work. In order to avoid too big PRs, we should refactor the ZooKeeper rolling update to not depend on the stateful set. It could add a new ZooKeeperRoller class which uses the same logic as we used so far but instead of depending on stateful sets it uses pod selector to find the pods and roll if needed. It also removed the rolling update logic from the StatefulSet operator and its Kafka and Zoo version since neither one uses it anymore.