[Inference/Refactor] Refactor compilation mechanism and unified multi hw (#5613)

* refactor compilation mechanism and unified multi hw

* fix file path bug

* add init.py to make pybind a module to avoid relative path error caused by softlink

* delete duplicated micros

* fix micros bug in gcc

extensions/csrc/kernel/cuda/utils/micros.h

@@ -0,0 +1,18 @@
+#pragma once
+
+#include <cuda.h>
+#include <cuda_runtime.h>
+
+#include <exception>
+
+#define CUDA_CHECK(func)                                    \
+  {                                                         \
+    auto status = func;                                     \
+    if (status != cudaSuccess) {                            \
+      throw std::runtime_error(cudaGetErrorString(status)); \
+    }                                                       \
+  }
+
+#define HOST __host__
+#define DEVICE __device__
+#define HOSTDEVICE __host__ __device__