Numpy 2.0 compatibility

[bmcfee]

This is a place-holder issue to verify compatibility with numpy 2.0.

At present, we're unable to evaluate this due to intermediate dependency incompatibilities (sklearn, numba, soundfile, etc). But as soon as those are sufficiently resolved, we should get on with verifying that everything works.

Linking upstream issues here so we can keep track:

• Ecosystem compatibility with numpy 2.0 numpy/numpy#26191 tracks general ecosystem compatibility and minimum versions

From what's currently listed there, we're looking at the following:

• ☑️ scipy 1.13.0
• ☑️ matplotlib 3.8.4
• ☑️ scikit-learn 1.4.2 Numpy 2.0 release thoughts scikit-learn/scikit-learn#27075
• ❌ numba 0.60 (not yet released) Numpy 2.0 binary support testing numba/numba#9466
• ❌ soundfile ??? (but I expect this to work)
• ❌ soxr-python ??? (likely needs a new build)

[bmcfee]

I'm tagging this onto a 0.11.0 milestone, which we may or may not do before 1.0. I'm leaving the possibility open in case we do need to change anything to support numpy 2.0 in a breaking way. I can imagine there's potential for stacking up a bunch of other version bumps (including Python), and that seems appropriate for a milestone release if it takes substantial effort to get a working environment spec.

[glemaitre]

FYI: scikit-learn 1.4.2 on PyPI should be compatible and was released.

[bmcfee]

Excellent, thanks for the update!