introduce CUB in sgl-kernel (#2887)

.gitmodules

@@ -1,3 +1,6 @@
 [submodule "sgl-kernel/3rdparty/cutlass"]
 	path = sgl-kernel/3rdparty/cutlass
 	url = https://github.com/NVIDIA/cutlass.git
+[submodule "sgl-kernel/3rdparty/cub"]
+	path = sgl-kernel/3rdparty/cub
+	url = https://github.com/NVIDIA/cub.git

sgl-kernel/CMakeLists.txt

@@ -9,6 +9,7 @@ set(CMAKE_CUDA_STANDARD 17)
 set(CMAKE_CUDA_STANDARD_REQUIRED ON)
 
 set(CUTLASS_DIR "3rdparty/cutlass")
+set(CUB_DIR "3rdparty/cub")
 
 # Set CUDA architectures
 set(CMAKE_CUDA_ARCHITECTURES "75;80;86;89;90")
@@ -43,6 +44,7 @@ target_include_directories(_kernels
         ${TORCH_INCLUDE_DIRS}
         ${CUTLASS_DIR}/include
         ${CUTLASS_DIR}/tools/util/include
+        ${CUB_DIR}/cub
 )
 
 target_link_libraries(_kernels