CUDA: Use MSVC default pattern for naming object files

The default that CUDA uses causes failures when you try to embed
CUDA obj's into another target.

Tests/Cuda/ObjectLibrary/Conflicts/CMakeLists.txt

@@ -0,0 +1,2 @@
+
+add_library(CudaConflicts OBJECT static.cu)

Tests/Cuda/ObjectLibrary/Conflicts/static.cu

@@ -0,0 +1,17 @@
+
+#include <cuda.h>
+#include <cuda_runtime.h>
+#include <iostream>
+
+int __host__ cu2_sq_func(int x)
+{
+  cudaError_t err;
+  int nDevices = 0;
+  err = cudaGetDeviceCount(&nDevices);
+  if (err != cudaSuccess) {
+    std::cerr << "nDevices: " << nDevices << std::endl;
+    std::cerr << "err: " << err << std::endl;
+    return 1;
+  }
+  return x * x;
+}