gcc 14.x broke nvcc

[totaam]

Looks like redefinition errors when using certain std headers in GMF.
Compiling the CUDA modules now fails (command extracted from running the build as usual):

$ CC=clang /opt/cuda-12.3.2/bin/nvcc -fatbin -std=c++17 -arch=all  -Wno-deprecated-gpu-targets -Xnvlink -ignore-host-info --allow-unsupported-compiler -c fs/share/xpra/cuda/RGBA_to_RGBAP.cu -o fs/share/xpra/cuda/RGBA_to_RGBAP.fatbin
/usr/include/c++/14/x86_64-redhat-linux/bits/c++config.h(2668): error: user-defined literal operator not found
    typedef __decltype(0.0bf16) __bfloat16_t;
                       ^
/usr/include/c++/14/type_traits(529): error: type name is not allowed
      : public __bool_constant<__is_array(_Tp)>
                                          ^
/usr/include/c++/14/type_traits(529): error: identifier "__is_array" is undefined
      : public __bool_constant<__is_array(_Tp)>
                               ^
/usr/include/c++/14/type_traits(581): error: type name is not allowed
      : public __bool_constant<__is_member_object_pointer(_Tp)>
                                                          ^
(...)
20 errors detected in the compilation of "fs/share/xpra/cuda/RGBA_to_RGBAP.cu".

That's with: gcc (GCC) 14.1.1 20240620 (Red Hat 14.1.1-6) from Fedora 40.

Adding -ccbin=`which clang++` as a workaround seems to work.
This is going to require packaging updates.

The MS Windows builds are not affected because on that platform nvcc uses MSVC instead. (the current version of gcc in MSYS2 is 14.1.0)