refactor: 统一硬件相关头文件引用

将分散在各文件中的CUDA/HIP/MUSA硬件相关头文件引用统一到vendors目录下的对应头文件中，提高代码可维护性。移除重复的头文件引用，优化构建配置。

csrc/quantization/gptq/matrix_view.cuh

@@ -6,8 +6,8 @@ https://github.com/turboderp/exllama
 #ifndef _matrix_view_cuh
 #define _matrix_view_cuh
 
-#include <cuda_runtime.h>
-#include <cuda_fp16.h>
+#include "../../vendors/functions.h"
+
 
 #include "qdq_util.cuh"
 

csrc/quantization/gptq/q_gemm.cu

@@ -6,11 +6,8 @@ https://github.com/qwopqwop200/GPTQ-for-LLaMa
 #include <cstdint>
 #include <cstdio>
 
-#include <torch/all.h>
-#include <c10/cuda/CUDAGuard.h>
-#include <ATen/cuda/CUDAContext.h>
-#include <cuda_runtime.h>
-#include <cuda_fp16.h>
+#include "../../vendors/functions.h"
+
 
 #include "compat.cuh"
 #include "matrix_view.cuh"