ENH: add parameter strict to assert_allclose

[mdhaber]

This PR adds the parameter strict to np.testing.assert_allclose. With strict=True, the shape and dtype of the two arguments must match for the assertion to pass.

Follows the example of gh-21595
Partially addresses suggestion in #24667 (comment)

[mdhaber]

To fix the CircleCI failure.

[mdhaber]

IIUC, I should add a file doc/release/upcoming_changes/24680.new_feature.rst with contents like

``strict`` option for `testing.assert_allclose`
-----------------------------------------------
The ``strict`` option is now available for `testing.assert_allclose`.
Setting ``strict=True`` will disable the broadcasting behaviour for scalars and
ensure that input arrays have the same data type.

? I'll also commit the <BLANKLINE>s above and run CI with [skip actions] [skip cirrus] [skip azp]?

[seberg]

That seems about right, yes.

[mdhaber]

Two minor tweaks for the reviewer to consider. Feel free to commit.
When this looks good, I'll submit a follow-up for assert_equal, if that would be helpful?
These will help address #24667 (comment).

[mattip]

@mdhaber Please go ahead and commit your self-review changes so we can get this in.

[mdhaber]

Thanks @mattip. That rendered as intended.
When this merges, shall I do the same for assert_equal?

[mattip]

Thanks @mdhaber. Yes, I think assert_equal could also get a strict kwarg. At the same time, the docstring of assert_equal could use some touchup: it is missing the Notes and See Also that the others have.

[mdhaber]

@mattip I opened gh-24770 to add the strict kwarg to assert_equal and touch up the docs as you suggested. gh-24775 adds strict to assert_array_less.