fix: re-order cases to handle NumPy scalar types properly

[agoose77]

#2096 improved our type-handling, but had a sneaky bug in the ordering of the NumPy scalar type (np.int16()) handling: the generic Integral test would always catch these cases first, meaning that ak.type(np.int16) always produced ScalarType(NumpyType("int64")) etc.

This PR simplifies this logic by re-ordering the tests.

[codecov]

Codecov Report

Merging #2125 (04cdb9c) into main (1e22929) will increase coverage by 0.00%.
The diff coverage is 54.16%.

Additional details and impacted files

Impacted Files	Coverage Δ
src/awkward/_connect/rdataframe/from_rdataframe.py	0.00% <ø> (ø)
src/awkward/contents/bitmaskedarray.py	68.02% <0.00%> (ø)
src/awkward/contents/bytemaskedarray.py	88.80% <0.00%> (ø)
src/awkward/contents/listarray.py	90.08% <0.00%> (ø)
src/awkward/contents/listoffsetarray.py	80.11% <0.00%> (ø)
src/awkward/contents/numpyarray.py	91.37% <0.00%> (ø)
src/awkward/contents/recordarray.py	84.05% <0.00%> (ø)
src/awkward/contents/regulararray.py	88.47% <0.00%> (ø)
src/awkward/contents/unionarray.py	84.98% <0.00%> (ø)
src/awkward/types/type.py	95.95% <ø> (ø)
... and 2 more

[jpivarski]

See below.