Previously we ran a HEAD against / and any 2XX or 3XX response would
indicate the server was up. Our new image builds coincide with gitea
server restarts and we've discovered that image builds are getting 503
errors requesting git repo data but presumably HEAD / is returning a 2XX
during this restart period. Make the health check more robust by
checking the actual healthz endpoint which checks the server status
internally instead. This should hopefully keep servers out of rotation
when they are not ready for requests.
Change-Id: I2c6c23027713ebe81d5e07e641759fd77ad01bb9
790fa8bfc2