Health endpoint indicating the node is ready

[alenkacz]

To be able to guarantee that availability is not affected, we need to be sure redpanda node rejoined raft groups so we can move on to another node (during update/upgrade procedure). This is particularly important for the kubernetes deployments.

We need an endpoint that will return 200 - 400 HTTP code when the node is ready.

Some more context from slack discussion with @dimitriscruz

A proposal would be that the operator calls an endpoint like http://broker1:9644/health/in-sync  to verify the target broker is in-sync with other replicas across all of its groups. The return type could be boolean or perhaps short json.
Another example is health/decommissioned.
Then we could also have http://localhost:9644/health to retrieve raw metadata for cases requiring their own semantics that don't have an existing endpoint.
It's unclear if the k8s placeholders for probes fit in well with the upgrade case. (e

[joejulian]

Duplicate of #4086