numpy 2.0: workaround regression in rescaleData

[pijyoi]

Testing against NumPy 2.0rc1, a performance regression was noticed in functions.rescaleData.

The cause was traced to:
https://numpy.org/devdocs/numpy_2_0_migration_guide.html#changes-to-numpy-data-type-promotion

It does seem rather counter-intuitive that the new behavior should affect in-place operations:
e.g.

data -= offset
data *= scalar

i.e. the resultant dtype can't change, so the performance regression might imply that a temporary higher precision ndarray was created? This would defeat the purpose of doing in-place operations.

The fix in this PR is only applied to the numpy codepath, and not to the cupy and numba codepaths.
The numba codepath (functions_numba.rescaleData) already marks the offset and scale arguments as type f8.

[j9ac9k]

Thanks for keeping the library up to date w/ issues with upcoming releases of our dependencies!