Div by 256 -> shift

[eddiewastaken]

TODO notes slowness, so change divide by 256 to right shift by a byte. On x86_64 with gcc, logarithmic gains between 0 bit input (same speed) and 256 bit input (2.65x faster). 1.7x speedup on average

[jgriffiths]

Hi @eddiewastaken thanks for contributing.

I think this change would only be valid if num was an intrinsic type (and the compiler didn't already optimize power of 2 division to shifts). But in this case the type is a boost 256 bit unsigned number parsed from a string, so the optimization doesn't apply (subject to the implementation of the class in boost).

However, it turns out that ga_session::sign_challenge is now unused since the login refactoring in 6222e42 and its follow up work. So, if you'd like to update this PR to remove the sign_challenge method, the uint256_to_base256 call and the boost/multiprecision/cpp_int.hpp include, I'd be happy to merge that!

[eddiewastaken]

Cool! Done as suggested, and raised with Boost :)