Allow parallel transfer_alm

[zatkins2]

pixell/cython/cmisc.pyx

Line 121 in 3906777

# I have C versions of this, but this one is more flexible and should be fast enough,

As discussed, in restricted situations, it would be nice if this function fell back to the parallelized C version (from slack):

Sigurd Naess
the cython version supports non-contiguous arrays and the op argument
I guess I could make it call the C version if the matrices are contiguous, and op is not passed