Add KV-cache quantization (kv_bits 8|4), golden-gated against mlx-lm#34

hvasconcelos merged 1 commit into

masterhvasconcelos/libmlxforge:masterfrom

kv-cache-quantizationhvasconcelos/libmlxforge:kv-cache-quantizationCopy head branch name to clipboard