For instance, when an user tries to import an image in the help document, assume it's an error and make a new document with the image.

gitlab-ci: now that we use Gnome runners, check and Linux tests can be run even by contributors on their repo

Seems to happen sometimes on Windows. As far as I see, not much can be done,
but assuming the page is empty should then trigger the OCR.
See Openpaper.work's bug report #468.

It returns a box, not just the text size.

--> do not crash ; replace the image by a placeholder

# Conflicts:
#	paperwork-gtk/src/paperwork_gtk/model/help/data/usage.tex

Core: Tests: 'paperwork-xxx plugins add/remove': do not mock the plugin 'config.backend.configparser'. Instead test it also

Now it's available, and performances are much better than loading a widget tree for each document row