forcing compilers to inline (or to not inline)

[Aleks Peshkov]

I discovered the artifact of inlining unlimited recursion by observed speed up and profiler output. It was a valid and complete perft engine. Later changes broke unrolling and I am not sure where was the difference.

MSVC2005 was not able to keep m_Lock inside a register.

GCC seemed to be worse in register optimization, but it is much better understanding inline assembler due extended syntax.