Optimize _rotate_basis_state

[emerali]

• Create cplx-torch version of np.prod to speed up GPU run-time
• Low prio: turn it into a generator function in order to be able to deal with a larger number of local rotations

[emerali]

Tried creating a cplx-torch version of np.prod, didn't actually improve GPU performance (in fact it made it slightly slower)

Implementation of cplx.prod was as so:

• Calculate arg of every element and add them up
• Calculate magnitude and multiply them
• Create complex result using magnitude and arg

There may be a better way to implement it, so I'll leave this issue open for now