test_util: avoid overflow errors in NumPy 2.0

[jakevdp]

Part of #19246.

NumPy 2.0 added OverflowErrors on array creation if the input is not representable under the requested dtype. For example:

>>> np.array(1000, dtype='uint8')
---------------------------------------------------------------------------
OverflowError                             Traceback (most recent call last)
Cell In[2], line 1
----> 1 np.array(1000, dtype='uint8')

OverflowError: Python integer 1000 out of bounds for uint8

This can be prevented by explicitly calling astype:

>>> np.array(1000).astype('uint8')
array(232, dtype=uint8)

This PR fixes a number of test failures under NumPy 2.0 that are related to this NumPy 2.0 behavior.