ffmpeg: use unwrapped clang for cuda-llvm support

The build system retargets clang for the appropriate nvidia arch
and doesn't depend on anything from the wrappers. Since the wrappped
compiler sets the flags for the original arch, it now includes
`-mtls-dialect=gnu2` on x86_64, which is not a supported flag for the
retargeted clang.

pkgs/development/libraries/ffmpeg/generic.nix

@@ -841,7 +841,7 @@ stdenv.mkDerivation (
     ++ optionals stdenv.hostPlatform.isx86 [ nasm ]
     # Texinfo version 7.1 introduced breaking changes, which older versions of ffmpeg do not handle.
     ++ optionals (lib.versionAtLeast version "6") [ texinfo ]
-    ++ optionals withCudaLLVM [ clang ]
+    ++ optionals withCudaLLVM [ clang.cc ]
     ++ optionals withCudaNVCC [ cuda_nvcc ];
 
     buildInputs =