Numpy surprises

New in numpy version 2: numpy int types aren't promoted to bigger types.

>>> np.uint8(1) + 255
<python-input-21>:1: RuntimeWarning: overflow encountered in scalar add
np.uint8(0)
#
>>> np.__version__
'2.2.4'

The old numpy 1.24 used to promote numpy int types to 64-bit:

>>> type(np.uint8(1) + 1)
<class 'numpy.int64'>
>>> np.__version__
'1.24.2'