CI: Add CUDA 13.2 to relevant workflows

[ptheywood]

Adds CUDA 13.2 to relevant CI workflows

Closes #1364

[ptheywood]

CUDA 13.2 on windows is hitting some deprecation warnings being promoted to errors in Thrust, which are not being triggered under linux CI

       "D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj" (default target) (1) ->
       "D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj" (CudaBuildCore target) (1:49) ->
       (CudaBuildCore target) -> 
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/insertion_sort.h(93): error #1444-D: type "thrust::_V_300200_SM_750::less<float>" was declared deprecated ("Use cuda::std::less instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::less<float>" was declared deprecated ("Use cuda::std::less instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::less<float>" was declared deprecated ("Use cuda::std::less instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/insertion_sort.h(93): error #1444-D: type "thrust::_V_300200_SM_750::greater<float>" was declared deprecated ("Use cuda::std::greater instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::greater<float>" was declared deprecated ("Use cuda::std::greater instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::greater<float>" was declared deprecated ("Use cuda::std::greater instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/insertion_sort.h(93): error #1444-D: type "thrust::_V_300200_SM_750::less<int>" was declared deprecated ("Use cuda::std::less instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::less<int>" was declared deprecated ("Use cuda::std::less instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::less<int>" was declared deprecated ("Use cuda::std::less instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/insertion_sort.h(93): error #1444-D: type "thrust::_V_300200_SM_750::greater<int>" was declared deprecated ("Use cuda::std::greater instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::greater<int>" was declared deprecated ("Use cuda::std::greater instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]
         C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.2/include/cccl\thrust/system/detail/sequential/merge.h(99): error #1444-D: type "thrust::_V_300200_SM_750::greater<int>" was declared deprecated ("Use cuda::std::greater instead") [D:\a\FLAMEGPU2\FLAMEGPU2\build\tests\tests.vcxproj]

thrust::greater appears to have been marked as deprecated in CCCL 3.1 .

Need to check that cuda::std::greater etc available for all CUDA versions we support (/ all CCCL we support) else we'll need to make this change conditional

Edit: This will likely be a conflcit with #1367, where rocthrust/hipcub are used rather than a CCCL equivalent 😢

[ptheywood]

With CUDA 13.2, and HIP in-mind, switching thrust::less<T> and thrust::greater<T> to std::less<> and std::greater<> appears to beahave, atleast under linux with GCC 15.

This will need testing (i.e. run the test suite) with our minimum supported CUDA & host compiler (12.0, GCC10 + 12.4 MSVC 17) on all platforms to ensure that nvcc handles it correctly. I think this should be fine for CUDA 11.0+ from googling, but not 100% confident of that.

If that doesn't work, we'll need to use cuda::std for nvidia and then use ifdef soup for HIP support.