CsPspEmu: Some improvements

I am working on supporting x64 right now (with atrac3+ support on windows) and improving the compiler performance. Also I have improved the memcpy, memset emulator is using internally making it much faster (almost 10x). Also I have been working on improving lwl/lwr instructions that allow to read unaligned words. https://github.com/soywiz/cspspemu/issues/42

Results are awesome. And make lot of compilerPerf tests 10x faster than a real PSP.

You can see the results here (running in 64-bits and with the new lwl/lwr improvement):
64-bits fixed the LWC1 performance problem on Intel platforms.

Still lots of things to improve:

  • Local variable usage
  • Improve function calling