Wishlist: TurboQuant support?

[QROST]

https://research.google/blog/turboquant-redefining-ai-efficiency-with-extreme-compression/

Someone already trying it with promising ram saving:
https://www.reddit.com/r/LocalLLaMA/comments/1s36vnk/looking_for_feedback_porting_googles_turboquant/

[QROST]

I will move it from issue to discussion.

[QROST]

Attempt closing