Support for automatic node pool upgrades?

[ligfx]

Issue to collect thoughts and best practices for handling automatic node pool upgrades.

In GKE, nodepool upgrades are handled via one of two strategies: surge or blue-green. It doesn't seem like either of these will play well with the Redpanda Operator, which currently only decommissions brokers when the replica count changes. The old node(s) and pods will be destroyed, then the pods will get scheduled on the new nodes (I think?), and the cluster will have a perpetually missing broker until it's manually force-decommissioned.

Presumably EKS and AKS have similar functionality.

JIRA Link: K8S-212