[CUDA] Mark CUDA as a no-errno platform.

[jlebar]

Summary

CUDA doesn't support errno at all, so this is the right thing -- or at
least, in the right direction.

But also, this unbreaks the CUDA test-suite math tests [0] after D39481.
cuda_cmath.h forwards nexttoward() to builtin_nexttoward, which,
after that patch, was lowering to a libcall that doesn't exist in NVPTX.

[0] http://lab.llvm.org:8011/builders/clang-cuda-build/builds/14999

cfe/trunk/lib/Driver/ToolChains/Cuda.h

public:
   // ptxas.
   bool useIntegratedAs() const override { return false; }
   bool isCrossCompiling() const override { return true; }
   bool isPICDefault() const override { return false; }
   bool isPIEDefault() const override { return false; }
   bool isPICDefaultForced() const override { return false; }
   bool SupportsProfiling() const override { return false; }
   bool SupportsObjCGC() const override { return false; }
+  bool IsMathErrnoDefault() const override { return false; }
 
   void AddCudaIncludeArgs(const llvm::opt::ArgList &DriverArgs,
                           llvm::opt::ArgStringList &CC1Args) const override;
 
   void addClangWarningOptions(llvm::opt::ArgStringList &CC1Args) const override;
   CXXStdlibType GetCXXStdlibType(const llvm::opt::ArgList &Args) const override;
   void
   AddClangSystemIncludeArgs(const llvm::opt::ArgList &DriverArgs,