ci: Test all Python versions on Windows on scheduled runs

[ichard26]

Cleaner version of #13014. See #13016 for why we decided to only run Windows CI across all Python versions on a scheduled (currently weekly) basis over all of the time.

Closes #13011.

[ichard26]

And now all the Windows jobs are being disabled. Great. I'll take a look later.

[ichard26]

OK, this is somewhat convoluted, but it does allow us to avoid duplicating the entire Windows job definition. I pulled the idea from https://github.com/orgs/community/discussions/26253#discussioncomment-3250989.

Alternatively, if this is too convoluted, we could generate the Windows matrix dynamically in a preceding job and use that to configure the Windows jobs, like https://github.com/FFY00/arch-python-repo/blob/main/.github/workflows/build.yml.

[ichard26]

I tested the syntax in a separate repository:

• Full Windows matrix (it's supposed to be a scheduled run, but that would be hard to test, so I used a manual run as the trigger) https://github.com/ichard26/random/actions/runs/13275847157
• Normal Windows matrix (like on push or PR) https://github.com/ichard26/random/actions/runs/13275838655

[ichard26]

@sbidoul how do you do feel about this approach?

[sbidoul]

Sounds good. Do you know who will be notified of failures of the scheduled jobs?

[ichard26]

I remember being sent a GitHub notification when a workflow failed on a scheduled run, but let's consult the GHA documentation:

Notifications for scheduled workflows are sent to the user who initially created the workflow. If a different user updates the cron syntax in the workflow file, subsequent notifications will be sent to that user instead. If a scheduled workflow is disabled and then re-enabled, notifications will be sent to the user who re-enabled the workflow rather than the user who last modified the cron syntax.

Wow, this is not great. It's probably best to create an issue if the scheduled run fails... I'd rather not use a third-party action for this, so I'll try using the gh tool.

[ichard26]

Also, CI is still too flaky (due to #13153) to open a new issue every time a scheduled run fails. This means this is blocked until that issue is fixed.