Changes for version 10.0.0

[bbayles]

This PR updates the library for version 10.0.0.

What's new in this PR:

• Python 3.7 is deprecated
• Issue Improve math.isqrt() emulation #732 is fixed by dropping support for Python 3.7, which doesn't have math.isqrt
• Issue Have matmul() rely on private versions of batched() and transpose() #733 is fixed by un-deprecating batched and making it and matmul yield tuples instead of lists

There are several other feature updates; see the changelog for details.

[rhettinger]

These edits look like nice improvements. Thank you.

Also consider using the older version of the convolve() recipe. It is faster than the currently published version which has more educational aims (explaining the algorithm for those who don't know it already and demonstrating how sliding_window() can be used) rather than speed. Besides inlining sliding_window() the older version avoids an unnecessary cast to tuple() and runs sumprod() directly on the deque instance.

def convolve(signal, kernel):
    # See:  https://betterexplained.com/articles/intuitive-convolution/
    # convolve(data, [0.25, 0.25, 0.25, 0.25]) --> Moving average (blur)
    # convolve(data, [1, -1]) --> 1st finite difference (1st derivative)
    # convolve(data, [1, -2, 1]) --> 2nd finite difference (2nd derivative)
    kernel = tuple(kernel)[::-1]
    n = len(kernel)
    window = collections.deque([0], maxlen=n) * n
    for x in chain(signal, repeat(0, n-1)):
        window.append(x)
        yield sumprod(kernel, window)

[bbayles]

Thanks - I've update the convolve recipe.

I'll publish version 10.0.0 next week. Given how widely deployed this package is, I want to make sure pushing a major version doesn't cause problems for people with less-than-ideal CI system configurations just before the weekend.