Release: NP2 Overflow Hotfix

[mikedh]

• Numpy 2 has stricter rules about overflowing integers, in an upstream application I was seeing grouping.hashable_rows overflowing an integer.
  • fixed by always promoting float_to_int to an int64 on integer input rather than returning any integer type. I also disabled the "int32 vs int64" choice based on maximum value as it may have provided a speedup in some cases, but making this function simpler and have consistent return types seems more valuable. The type hint will be checked by Beartype in the docker image tests.
• Removed the prerelease install for Numpy 2 in tests since Numpy 2.0.0 was released June 16.

Traceback for posterity:

File "/home/mikedh/trimesh/trimesh/visual/color.py", line 431, in main_color
    unique, inverse = unique_rows(colors)
                      ^^^^^^^^^^^^^^^^^^^
  File "/home/mikedh/trimesh/trimesh/grouping.py", line 451, in unique_rows
    rows = hashable_rows(data, digits=digits)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/mikedh/trimesh/trimesh/grouping.py", line 209, in hashable_rows
    bitbang = (as_int.T + threshold).astype(np.uint64)
               ~~~~~~~~~^~~~~~~~~~~
OverflowError: Python integer 32767 out of bounds for uint8