Try building with numpy 2.0 rc1

[takluyver]

See discussion on #2353.

[takluyver]

Numpy 2.0 requires Python >= 3.9, whereas we're still testing against 3.8. So for now I've made it conditional: it continues using oldest-supported-numpy to build on Python 3.8.

We can discuss dropping Python 3.8 as a separate question. The NEP 29 recommendation is to drop it later this month, though I tend to err on the cautious side with old Python versions, not least as we're still using 3.8 in some places at XFEL.

[takluyver]

Looks like we've got some issues on Windows, mostly similar to:

>           assert normalized_dset_dtype == normalized_arr_dtype, \
                "Dtype mismatch (%s vs %s)%s" % (normalized_dset_dtype, normalized_dset_dtype, message)
E           AssertionError: Dtype mismatch (int64 vs int64)

[takluyver]

Ah, I see there's a bug in the helper function that makes the message less than helpful.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.53%. Comparing base (078a481) to head (e366e49).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2401   +/-   ##
=======================================
  Coverage   89.53%   89.53%           
=======================================
  Files          17       17           
  Lines        2380     2380           
=======================================
  Hits         2131     2131           
  Misses        249      249           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[takluyver]

That seems to be working; I'll merge in another day or so if there are no objections.

[aragilar]

Looks good to me, and agree on keeping python 3.8 supported for now.