Add TurboQuant KV cache compression (3-bit, 4.6x)#1067

arozanov wants to merge 5 commits intoml-explore:mainml-explore/mlx-lm:mainfrom

arozanov:feature/turboquant-kv-cachearozanov/mlx-lm:feature/turboquant-kv-cacheCopy head branch name to clipboard