HDFS controller doesn't watch Listeners

[nightkr]

Controllers should watch for every object that they read, otherwise it's easy to end up "stuck" with stale data.

In this case, NameNode listeners are used to feed the discovery configmap, so we should reconcile whenever they change.

The issue is somewhat mitigated by that pod startup is inhibited by listener-operator until an address has been acquired (which we do watch, via the StatefulSets), but it's still a good to paint both sides of the fence...