BUG: Check whether the datatype is of Datetime before typecasting as usual.

[Dinkarkumar]

Closes #19267, np.allclose fails with dtype datetime64

This is my first contribution, Seeking guidance and support from such huge community.

[mattip]

Please add a test that fails before, passes after the change.

[Dinkarkumar]

Reproducing code example:
Earlier the following code snippet used to give TypeError:-
import numpy as np
a = np.zeros(1, dtype='datetime64[ns]')
np.allclose(a, a)

Now after code changes the typecasting would not be possible and the np.allclose(a,a) would run for even dtype = 'datetime64[ns]'
@mattip , i am not yet convinced of why do you ask me to revert the chnanges. Let's discuss and resolve the issue.

[mattip]

I asked only that you revert the changes to README.md. You still did not add a test, the place to add it would be numpy/core/tests/test_numeric.py.

[Dinkarkumar]

Sorry for such immature execution of my task. I am totally new to this environment. Hope to learn and seeking guidance from you all.

[mattip]

I think you might want to back out all the changesets other than abe9ff0 and then add the test in a second changeset. Your PR should only consist of

• a one-line change to numpy/core/numeric.py
• a test added/modified in numpy/core/tests/test_numeric.py, which should also be at the most 10 lines.

[seberg]

I am going to close this. The fix is likely right, but its a mess unfortunately and old. It also needs a rebase anyway, creating a new PR from scratch seems just as well (which anyone is very welcome to do!).