Optimization memory allocation NUMA-awareness

[GermanAizek]

I would like to suggest a feature once we touched NUMA #4167,
in NUMA you can optimize memory allocation, namely, allocate it for each logical group of processors and touch memory, thus you can speed up by 2 times with 2 cpu, x4 cpu * 4 theotically.

I have already done similar implementation, but only on WinAPI in x64dbg project here: x64dbg/x64dbg#3307

I have made support in function both platforms singleCPU and NUMA.

In reality, rewriting VirtualAlloc on VirtualAllocExNuma resulted to 1.5-fold increase in execution find xrefs process.