Largely cosmetic change - add two new template params to cpp_dec_float - one for the exponent type, and one for an optional allocator.