Cleanup CI system

[jarrodmillman]

There are still some flaky CI tests. For example, see https://github.com/scikit-image/scikit-image/actions/runs/12977550163/job/36191096309

Also there are several warnings which should be addressed or silenced.

We should also re-enable code coverage.

[lagru]

Also there are several warnings which should be addressed or silenced.

I guess these have accumulated over time because our CI hasn't been failing for warnings in our test suite (see #7554). The ones I see in your linked job look benign. Most is from the deprecation of skimage.io, which should be ignored once pyproject.toml is taken into account; there's a warning filter in

scikit-image/pyproject.toml

Line 197 in bbbfcdb

"ignore:.*use `imageio` or other I/O packages directly.*:FutureWarning:skimage",

.

The other ones, I'll naturally have to address for #7554 as well, because they would otherwise result in errors.

I'll try to get this ready for review soon.

[lagru]

We should also re-enable code coverage.

We discussed this right now with @stefanv and @matthew-brett. It didn't sound like it would be a high priority right now but it might be worth it with a setup that makes a PRs impact on coverage very transparent.

I remember that we disabled coverage originally because nobody was really looking at the numbers reported in the CI logs. It was just causing our CI to run slower.

[jarrodmillman]

The warnings aren't a big deal,but they make scrolling through the test logs noisy. So cleaning it up would be helpful, but mostly for developers reviewing PRs. Thanks for taking this on.