Use standard attention for Ministral3 (#1032)

Required adding the "temperature scaling" to the standard attention
implementation.

But in this way split mode "graph" is automatically supported.

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

src/llama.cpp

@@ -1728,6 +1728,7 @@ static bool is_model_split_supported(const llama_model & model) {
         LLM_ARCH_LLAMA,
         LLM_ARCH_QWEN3MOE,
         LLM_ARCH_GLM4_MOE,
+        LLM_ARCH_MISTRAL3,
     };
     auto it =  k_supported.find(model.arch);
     return it != k_supported.end();