fix return type of cuBLAS (#749)

Author: zhyncs

csrc/activation.cu

@@ -15,7 +15,7 @@
  */
 #include <flashinfer/activation.cuh>
 
-#include "aot_extension_utils.h"
+#include "pytorch_extension_utils.h"
 
 using namespace flashinfer;
 

include/flashinfer/gemm/bmm_fp8.cuh

@@ -45,7 +45,7 @@
 #else
 #define FLASHINFER_CUBLAS_CALL(EXPR)  \
   {                                   \
-    cudaError_t e = (EXPR);           \
+    cublasStatus_t e = (EXPR);        \
     if (e != CUBLAS_STATUS_SUCCESS) { \
       return e;                       \
     }                                 \