Completely remove mont2 stuff.
It does not appear to be faster than the current Montgomery code
except for very small moduli (somewhere between 192 and 224 bits
in a 64-bit Sun environment, and even less than 192 bits
on 32 bit systems).