Rm multiplication using NAF

Comb method is at most 1% slower for random points,
and is way faster for fixed point (repeated).
2 files changed