Add's workaround for ScalarType::Byte for cuda (pytorch#35027)

Summary:
This PR add's a workaround for `cuda` for `ScalarType::Byte` for the `AT_DISPATCH_*` macros.
As discussed here:
pytorch#34826
Pull Request resolved: pytorch#35027

Differential Revision: D20596555

Pulled By: colesbury

fbshipit-source-id: 72e842603723a5aa146e4224e79befafc62f2624

Author: r-zenine

c10/core/ScalarType.h

@@ -108,6 +108,17 @@ struct ScalarTypeToCPPType<c10::ScalarType::Bool> {
   static type t;
 };
 
+template<>
+struct ScalarTypeToCPPType<c10::ScalarType::Byte> {
+  using type = uint8_t;
+
+  // This is a workaround for the CUDA bug which prevents ::detail::ScalarTypeToCType<T>::type being used directly
+  // due to ambiguous reference which can't to be resolved. For some reason it cant pick between at::detail and at::cuda::detail.
+  // For repro example, please see: https://gist.github.com/izdeby/952ae7cf256ddb740a73776d39a7e7ba
+  // TODO: remove once the bug is fixed.
+  static type t;
+};
+
 template<>
 struct ScalarTypeToCPPType<c10::ScalarType::Long> {
   using type = int64_t;