Rpcs3: RSX memory copy optimization

[sampletext32]

https://github.com/RPCS3/rpcs3/blob/cce946f10e809319020a2c2559b0f3ab9af5a527/rpcs3/Emu/RSX/rsx_methods.cpp#L438-L441

@kd-11, @elad335 , I think this area can be optimized with custom mem copy.

Explanation.

Now we have mem check and after that mem copy. On very low level it is 2 runs over memory segment. But if we copy values without check (just remove memcmp) we can achieve the same result. We can assign "m_graphics_state" flag on the run with condition. On very low level it would be represented as 1 run over memory segment. This approach decreases amount of cycle bounds checks, and doesn't affect any actual data checks.

Maybe it's not worth it, but a tiny speedup can help the emulation.

[elad335]

Thanks, fixed by #7992, it also takes care stream_data_to_memory_swapped_u32 which makes it super fast now.