[Nvidia][SM121] Add intrin.h include to gemm_mma.h for sm120+(#785)

To make sm120 arch runnable.

src/tl_templates/cuda/gemm_mma.h

@@ -9,6 +9,7 @@
 
 #include "common.h"
 #include "cuda_fp8.h"
+#include "intrin.h"
 
 namespace cute {