deployer: health check can fail to detect failing startup

[consideRatio]

In #3776 I saw that the health check passed, but manually testing to launch a server failed. I assume the difference is that we check that the pod enters a running state or similar, but for the user server to successfully start - also JupyterHub needs to believe its successfully started, and that isn't something we check for currently.

Action points

• Look into how our health check test work, does it consider JupyterHub's understanding about the server status?
  • Note that a server can be considered started, but not ready to be redirected to as well - what happens in this case is that the server is started by KubeSpawner - but JupyterHub never redirects the user because its not considering the user server ready yet.

[consideRatio]

I think this could have been due to a kubespawner bug fixed in kubespawner 7, available in z2jh 4, so I propose ignoring this and relying on getting z2jh 4 (see #5079 )