Move kv cache scales from k/v_proj.output_scale to self_attn.k/v_scale#133

mgoin merged 2 commits intomainneuralmagic/compressed-tensors:mainfrom

move-kv_cache_scheme-to-kv_scalesneuralmagic/compressed-tensors:move-kv_cache_scheme-to-kv_scalesCopy head branch name to clipboard