CI fails when numpy 2.0.0 is installed

[orbeckst]

Expected behavior

CI passes with correct code.

Actual behavior

Azure CI runners such as https://dev.azure.com/mdanalysis/mdanalysis/_build/results?buildId=7066&view=logs&jobId=86224841-1158-5011-fbb7-7c2644c2081b&j=86224841-1158-5011-fbb7-7c2644c2081b&t=44ce4b91-ecdc-5aca-e807-60d46710cf6d in #4162 (comment) fail with

=================================== ERRORS ====================================
__________________ ERROR collecting MDAnalysisTests/analysis __________________
C:\hostedtoolcache\windows\Python\3.10.11\x64\lib\importlib\__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
    from .lib import log
C:\hostedtoolcache\windows\Python\3.10.11\x64\lib\site-packages\MDAnalysis\lib\__init__.py:34: in <module>
    from . import transformations
C:\hostedtoolcache\windows\Python\3.10.11\x64\lib\site-packages\MDAnalysis\lib\transformations.py:172: in <module>
    from .mdamath import angle as vecangle
C:\hostedtoolcache\windows\Python\3.10.11\x64\lib\site-packages\MDAnalysis\lib\mdamath.py:63: in <module>
    from . import util
C:\hostedtoolcache\windows\Python\3.10.11\x64\lib\site-packages\MDAnalysis\lib\util.py:228: in <module>
    from ._cutil import unique_int_1d
MDAnalysis\\lib\\_cutil.pyx:1: in init MDAnalysis.lib._cutil
    ???
E   ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

The action installed numpy 2.0.0 via pip.

[orbeckst]

@tylerjereddy @IAlibay any ideas what's happening and how to fix it?

[tylerjereddy]

Pretty sure pyproject.toml needs an update so we don't build wheels (against NumPy 1.x) that are binary incompatible against the NumPy 2.x series. I opened the cross-listed PR. As @rgommers indicated, we should try to get a 2.x compatible release out soon-ish too.

[tylerjereddy]

it looks like some of our Windows deps may not have NumPy 2 compatible releases yet as well