if bugzilla is unavailable during setup, do not fail the run if no tests that are collected require the bugzilla marker

[InbarRose]

when running pytest, if bugzilla is not responsive or not reachable (or even lets say there is missing configuration) one way or another, if the bugzilla process fails during the initialization it should not fail the pytest process if none of the tests that will be collected/run actually require the bugzilla marker.

this is especially frustrating if i specify a specific test to run which i know does not use the bugzilla marker, and the bugzilla setup fails and it will not run my test.

[petr-balogh]

Just wanted to open similar issues and found this.

For example yesterday, there was an outage and some of our execution just errored a lot of test cases in the setup so they event didn't try to run.

I would expect that if there is problem with bugzilla, that it will run the test as IMO it's better to have failed test cause of bug then do not run it at all and fail in it's setup.

This behavior can be optional based on some CLI argument like --bugzilla-ignore-error

[petr-balogh]

@els-pnw please do you have some thoughts on this issue?