enhancement: axes keyword for cross

[2sn]

It would be great to have an "axes" keyword in np.cross for consistent API to np.tensordot.

Reproducing code example:

import numpy as np
a = [1, 2, 3]
b = [4, 5, 6]
np.cross(a, b, axes = (-1, -1))

Currently, rather ugly axisa, axisb, axisc, or just axis have to be provided. I assume it could be added to overwrite the existing keywords for backward compatibility. I propose that the new axes keyword may be an iterable (or tuple) of length two or three, withe the third value defaulting to axisc or -1.

[WarrenWeckesser]

Related: #13718