PR workflow fixes

[jikortus]

• Add missing dependencies (even though they only relate to lanabot/runtime/installation/hub/keyboard/layouts.py, it seems better to add them instead of ignore the import errors in pylint)
• Ensure no containers using the base image are left before removing the image. This should address the current failures of the deployment phase in PR workflow, even though I have no idea how there could still be some leftover containers that are actually using the image.

[jstodola]

Couldn't this be done simply by podman image rm --force ... ?

[jikortus]

Yes, it should work the same, I think. I may try to resort to the force deletion, but I wonder why the image still doesn't get deleted by the current solution, despite it not being used by any container (as you can see in the workflow script output). @velezd, you know most likely much more about the GitHub workflows - is it possible that this is a 'feature' for some reason?

[jikortus]

@jstodola so I tried to just enforce the image removal as you suggested and it works fine (as you can see in the checks below). However, just FYI, I think that even if we didn't do any cleanup at all, it shouldn't actually cause any mess, because the workflows use a fresh VM every time, which makes sense after all:

Each workflow run executes in a fresh, newly-provisioned virtual machine.

[jstodola]

It looks good to me. We can remove the cleanup part if it keeps causing issues in the future.

[jikortus]

Thanks for the review!