Handle trying to read logs of a container that has not fully started

[olevski]

We have this happening in the logs once in a while:

We should try to handle that error and return None or something from the k8s client. Or handle the error in our centralized error handler.

[leafty]

Note that this has been fixed for session logs, so the error handling should probably be copied to handle build pod logs.

[leafty]

#1057 should have been fixed mhm