Support numpy2

[pp-mo]

It seems that the only remaining problem for numpy2 is that the update to the code introduced here in #578 was causing tests to fail, when actually run with numpy 2.

I had meant to keep that issue separate from the eccodes 2v38 one, but somehow I tripped over my own commits.

This PR fixes that, so when merged it should unblock #550 🤞

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.61%. Comparing base (11169cf) to head (da4792b).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #580   +/-   ##
=======================================
  Coverage   89.60%   89.61%           
=======================================
  Files           8        8           
  Lines        2473     2475    +2     
  Branches      420      420           
=======================================
+ Hits         2216     2218    +2     
  Misses        159      159           
  Partials       98       98           

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

[trexfeathers]

I tested this locally and it does indeed eliminate this error:

AssertionError: np.uint32(4294967295) != -1

But I still get these two errors:

TypeError: a class that defines __slots__ without defining __getstate__ cannot be pickled

[pp-mo]

But I still get these two errors:

TypeError: a class that defines __slots__ without defining __getstate__ cannot be pickled

Oh that's a bit weird : I didn't get that.
Also, I don't think that is a numpy2 thing, it's just generally a fact about pickling, so I can't see why anything here should be relevant to it, or the adoption of numpy2 for that matter.
So, where does this crop up ??

[trexfeathers]

So, where does this crop up ??

It's in the failures for #550

[trexfeathers]

But I still get these two errors:

TypeError: a class that defines __slots__ without defining __getstate__ cannot be pickled

This was proven to be caused by an unrelated Dask problem - SciTools/iris#6258