Fixes #23145 - handle strong params for compute attributes on failure

[iNecas]

See Redmine for description of the issue.

[theforeman-bot]

Issues: #23145

[iNecas]

Screenshot of the failure I'm talking about:

[tbrisker]

Code makes sense, didn't test though.
I hope removing the rescue won't lead to other exceptions blowing up, but in any case swallowing them silently is a Bad Idea &tm; and we should handle them properly.

[tbrisker]

@iNecas - What happens if an actual error is encountered when talking to the compute resource?
@xprazak2 - I think we don't need to block branching on this, but i guess it would be good to get into the RC so we test this works properly

[iNecas]

It depends on when the call is actually made, during the orchestration, the exception will be captured, as any other exception would be. When rendering, it's harder to capture, but that's an issue that is already in other places: just think about any other call that we don't rescue from (such as listing templates for select box or list of networks for NIC).

Btw. next thing I would like to do is to remove test_connection call from new_vm call, as it incredibly slows down the whole process, and I don't see any point of having it there. But keeping that outside of this PR

[iNecas]

Tests are 💚, this should be ready for merge.

[ares]

Tested, fixes the problem

Thanks @iNecas, merging