quantize: add option to automatically choose optimal quant types to reach a bpw target at lowest error

include/llama.h

@@ -360,9 +360,12 @@ extern "C" {
         bool pure;                            // quantize all tensors to the default type
         bool keep_split;                      // quantize to the same number of shards
         void * imatrix;                       // pointer to importance matrix data
+        void * activations;                   // pointer to activations data
         void * kv_overrides;                  // pointer to vector containing overrides
         void * tensor_types;                  // pointer to vector containing tensor types
         void * prune_layers;                  // pointer to vector containing layer indices to prune
+        float target_bpw;                     // target bits per weight (bpw)
+        bool precise_lambda;                  // use precise_lambda calculation - slow computation but very accurate
     } llama_model_quantize_params;
 
     typedef struct llama_logit_bias {