redpanda: changing replicas results in rolling restarts

[chrisseto]

As of chart version 5.9.2, changing statefulset.replicas will result in needless restarts of the redpanda cluster. This is due to redpanda.yaml containing a list of brokers that is computed from .replicas.

The schemaregistry and pandaproxy clients can be updated to point to localhost as they'll discover the rest of the cluster through kafka metadata.

Rpk is also configured with a list of brokers. I'm not certain about the best way to handle this one as we use rpk cluster health in the readiness probe.

JIRA Link: K8S-352