Add Numpy strategy for basic_indexes(shape)

[Zac-HD]

This issue is extracted from #1829, and depends on slices() from #1929.

MyGrad defines a strategy for basic_indexes, and we would like to bring a simpler version of it into the upstream hypothesis.extra.numpy.

Given an array shape, it generates valid indices for an array of that shape which conform to the basic (i.e. no-copy) indexing specification. Examples from this strategy shrink from indices that produce high-dimensional views of the input array, down to indices that select a single element from the array.

An initial implementation could work something like:

def basic_indexes(shape):
    # Validation logic for shape here - consider extracting a check_shape helper.
    return tuples(*(slices(s) for s in shape))

An optional extension would add an allow_ellipsis=True flag, which picks 0-2 dims to replace with ... - only a few to ensure that the resulting arrays are typically reasonably small compared to the input shape, but may still generate a slice with a length != the number of dimensions to index.

np.newaxis (aka None) is also supported to add size-1 dimensions in any position, but we would put this behind a allow_newaxis=False flag. As a disabled-by-default additional feature, we would prefer to receive a follow-up pull request to support newaxis after the initial version is complete and released to users.