Fix for numpy2.0 compatibility

[PGijsbers]

Numpy2.0 cleaned up their namespace, so we have to alias for the new version.

The problem is that this is not covered in CI (tested versions do not support numpy 2.0), and in general tests need to be updated to support more recent scikit-learn versions that support 2.0 (1.5, maybe 1.4?). I don't know if I really have the time for that right now. Options:

• Put a pin in the dependencies
• Accept this as a best-effort, and hope that the tests that don't fail due to scikit-learn incompatibility sufficiently cover the code to raise any numpy1/2 issues
• Wait until I find time to make tests support scikit-learn 1.5

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 28.20%. Comparing base (923b49d) to head (124e96f).

Additional details and impacted files

```diff @@ Coverage Diff @@ ## develop #1341 +/- ## =========================================== - Coverage 32.94% 28.20% -4.75% =========================================== Files 38 38 Lines 5254 5255 +1 =========================================== - Hits 1731 1482 -249 - Misses 3523 3773 +250 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.