Single node down shows other node as broken

[mishamo]

Steps to reproduce:

• Scale up cluster with 2 brokers.
• Run kafka health check on each node.
• Stop the first broker.
• Hit the broker healthcheck endpoint on the second broker.

Expected result:

• 200 imok

Actual result:

• 500 nook

Presumably this is related to the fact that the broker healthcheck topic is not replicated onto the first broker and the second broker sees it as unhealthy when the first is down. Perhaps the second broker's healthcheck should ignore the healthcheck topic of the first?

The cluster is also shown as red from the second broker and yellow from the first, which shows an inconsistent view of things. Perhaps the broker healthcheck topics should be filtered out when polling for cluster health.