Memory profile optimiziation

[geonove]

Parallelizes memory usage profile trials with OpenMP and made the allocator counter thread-local so trials can run concurrently. Added thread-safe stats updates and per-thread state/RNG in each memory profile implementation, enabling multi-core runs without races and cutting runtime drastically.

Tested on a M1 MacBook Pro and went from:

Cum Trials             : 4
Cum Time, ms           : 546898
Avg Time Per Trial, ms : 136724
Current time           : Sun Jan  4 16:15:32 2026
Est Time of Completion : Sat Apr 18 10:05:29 2026

to:

Cum Trials             : 4
Cum Time, ms           : 146067
Avg Time Per Trial, ms : 36516
Current time           : Sun Jan  4 16:23:55 2026
Est Time of Completion : Sun Feb  1 09:06:41 2026