Now that #22554 was merged in main, it would be great to find a a way to run a weekly scheduled job to run the scikit-learn main test on a CI worker with an NVidia GPU and CuPy.

In case of failure, it could create a report as dedicated issues as we do for other scheduled jobs:

• https://github.com/scikit-learn/scikit-learn/blob/main/maint_tools/update_tracking_issue.py

Maybe @betatim has a plan and connections to do that? ;)