Add faster FpMul implementation

[ecnerwala]

FpMul is the first in a family of BigNum templates which do arithmetic over Fp but allow for aliased results, avoiding the (expensive) BigLess check that residues are less than the modulus. FpMul is a merged multiply and mod, which accumulates overflows and only computes range checks when reducing. Other potentially useful operations include FpAdd, FpSub, FpInv, FpMulAdd, etc; each has slightly different overflow bit lengths, so care should be taken when implementing. Additionally BigMultModP can be implemented as FpMul followed by a BigLess check.

Also simplify some generic bigint witness computation.