Healthcheck should be performed at startup, not only on error #507
Description
Currently healthchecks are only performed on nodes that are dead before they are resurrected. This means that a user can deploy logstash, have it function, then have a node become temporarily unavailable only to discover that it won't be detected as up by logstash because the healthcheck URL is not one Logstash has perms to access.
All nodes should start in an 'unknown' state and transition to 'good' to make this behavior more consistent. Additionally, we should provide a helpful error in the situation where the healthcheck URL fails.