Fix occupancy calculations API on NVCC (#1941)

NVCC warned if you tried to use hipOccupancyMaxActiveBlocksPerMultiprocessor
because when passing in a device function pointer, "const void* func" was
insufficient to describe it accurately. Adding a C++ templated class type
definition for this function.

[ROCm/hip commit: 4128d68ed7]

projects/hip/include/hip/nvcc_detail/hip_runtime_api.h

@@ -1708,6 +1708,17 @@ inline static hipError_t hipLaunchCooperativeKernelMultiDevice(hipLaunchParams*
 
 #ifdef __CUDACC__
 
+template<class T>
+inline static hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(int* numBlocks,
+                                                                      T func,
+                                                                      int blockSize,
+                                                                      size_t dynamicSMemSize) {
+    cudaError_t cerror;
+    cerror =
+        cudaOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSMemSize);
+    return hipCUDAErrorTohipError(cerror);
+}
+
 template <class T>
 inline static hipError_t hipOccupancyMaxPotentialBlockSize(int* minGridSize, int* blockSize, T func,
                                                            size_t dynamicSMemSize = 0,