Rejig the bottom-level loops in internal_mul_* to use pointers instead

Rejig the bottom-level loops in internal_mul_* to use pointers instead
of array indices. You'd hope that compilers could automatically turn
the one representation into the other if it was faster to do so, but
apparently not: even on gcc -O3, this source transformation gains over
15% performance.

[originally from svn r9105]