performance on AMD and Windows

lbrycrd runs slower on (modern) AMD processors than it does on modern Intel processors, even with SHANI enabled on AMD. It may be that some of the hand-tweaked optimizations are specific to Intel. This should be explored.

lbrycrd runs slower on MS Windows than it does on Linux. I thought it was an issue with the cross compiler, but it's true even when compiling with MSVC. In fact, the cross compiler does better than MSVC.

No files, tests, or entry points are named. Start by reproducing the performance difference on modern AMD and Intel processors and on Windows and Linux, including SHANI, cross-compiler, and MSVC builds; done means identifying the relevant platform or optimization cause and documenting a focused next change.

Written by the indexing model from the issue text.