ENH: Add fast array equal function for indexers

[phofl]

• closes #xxxx (Replace xxxx with the GitHub issue number)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

The all equal case is as fast as numpy on 2 million entries, the early exist is significantly faster obviously

[lithomas1]

Scrolling through issues, it looks like this is basically #32339.
(Takeaway from there was that perf for non short-circuit case needed to be checked. It's pretty hard to beat numpy's SIMD extensions.)

[phofl]

Commented on this in the initial post. Both are running around 700micro seconds if the whole array is checked (2 million entries). Ours is actually a bit faster, but not by much.

Numpy is doing some things that are more expensive, e.g. allocating a Boolean array and then calling all on it. +

Edit: Looks like NumPy is faster with less elements, but not by that much:

10**6
%timeit np.array_equal(left, right)
269 µs ± 5.64 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)
%timeit array_equal_fast(left, right)
303 µs ± 2.44 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

and with 2 Million:

%timeit array_equal_fast(left, right)
701 µs ± 79.4 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)
%timeit np.array_equal(left, right)
729 µs ± 7.5 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

Edit2: I'd try using this only for indexers in CoW in the beginning. The advantage is, that if an indexer is equal, we avoid a copy which brings a big speedup. And if it's not equal, we should find different values early on.

[jorisvandenbossche]

I check locally a little bit, and using a memoryview (int[:]) instead of ndarray doesn't seem to give a significant speed up here (I seemed to remember that those could sometimes be faster)

[jorisvandenbossche]

Looks good to me. Also for me locally it more or less gives the exact same timing as np.array_equal (for fully equal arrays)

[lithomas1]

LGTM. Nice speedup!