REL: Prepare for the NumPy 1.23.3 release.

- Update mailmap.
- Update 1.23.3-notes.rst
- Create 1.23.3-changelog.rst.

* BUG, SIMD: Handle overflow errors

Overflows for remainder/divmod/fmod

* If a types minimum value is divided by -1, an overflow will be raised
  and result will be set to minimum
* Handle overflow and return 0 in case of mod related operations

* TST: New tests for overflow in division operations

* SIMD: Removed cvtozero

Co-authored-by: Rafael Cardoso Fernandes Sousa <rafaelcfsousa@ibm.com>

* TST: Removed eval | Fixed raises cases

* TST: Changed `raise` to `warns`

* Changed `raise` to `warns` and test for `RuntimeWarning`
* Added results check back

* TST: Add additional tests for division-by-zero and integer overflow

This introduces a helper to iterate through "interesting" array cases
that could maybe be used in other places.  Keep the other test intact,
it adds a check for mixed types (which is just casts currently, but
cannot hurt) and is otherwise thorough.

* MAINT: Remove nested NPY_UNLIKELY in division paths

I am not certain the unlikely cases make much sense to begin with,
but they are certainly not helpful within an unlikely block.

* TST: Add unsigned integers to integer divide-by-zero test

* BUG: Added missing `NAME` key to loop generator

* BUG, SIMD: Handle division overflow errors

* If a types minimum value is divided by -1, an overflow will be raised
  and result will be set to minimum

* TST: Modified tests to reflect new overflow

* Update numpy/core/src/umath/loops_arithmetic.dispatch.c.src

Co-authored-by: Rafael Sousa <90851201+rafaelcfsousa@users.noreply.github.com>

Co-authored-by: Rafael Cardoso Fernandes Sousa <rafaelcfsousa@ibm.com>
Co-authored-by: Sebastian Berg <sebastian@sipsolutions.net>
Co-authored-by: Rafael Sousa <90851201+rafaelcfsousa@users.noreply.github.com>

BUG: Fix the implementation of numpy.array_api.vecdot

TST: ensure ``np.equal.reduce`` raises a ``TypeError``

* Fix the implementation of numpy.array_api.vecdot

See https://data-apis.org/array-api/latest/API_specification/generated/signatures.linear_algebra_functions.vecdot.html

* Use moveaxis + matmul instead of einsum in vecdot

If we cache a promoted version of the loop, that promoted can mismatch
the correct one.  This ends up being rejected later in the legacy paths
(should be the only path currently used), but we should reject it here
(or in principle we could reject it after cache lookup, but we are fixing
up the operation DTypes here anyway, so we are looking at the signature).

A call sequence reproducing this directly is:

np.add(1, 2, signature=(bool, int, None))  # should fail

np.add(True, 2)  # A promoted loop
np.add(1, 2, signature=(bool, int, None))  # should still fail

Not that the errors differ, because the first one comes from the old
type resolution code and is currently less precise

TYP,BUG: Reduce argument validation in C-based ``__class_getitem__`` 

TST,BUG: Use fork context to fix MacOS savez test