Fixing cmake charm++ and charm4py build with cuda

[mayantaylor]

Previously, attempting to build with the cuda option using the cmake build resulted in the following error (on Delta):

> -- Found CUDAToolkit: /sw/spack/deltas11-2023-03/apps/linux-rhel8-zen3/gcc-11.4.0/cuda-11.8.0-vfixfmc/include (found version "11.8.89") 
> -- The CUDA compiler identification is GNU 11.4.0
> -- Detecting CUDA compiler ABI info
> -- Detecting CUDA compiler ABI info - failed
> -- Check for working CUDA compiler: /sw/spack/deltas11-2023-03/apps/linux-rhel8-zen3/gcc-11.4.0/cuda-11.8.0-vfixfmc/bin/nvcc
> -- Check for working CUDA compiler: /sw/spack/deltas11-2023-03/apps/linux-rhel8-zen3/gcc-11.4.0/cuda-11.8.0-vfixfmc/bin/nvcc - broken
> CMake Error at /usr/share/cmake/Modules/CMakeTestCUDACompiler.cmake:52 (message):
>   The CUDA compiler
> 
>     "/sw/spack/deltas11-2023-03/apps/linux-rhel8-zen3/gcc-11.4.0/cuda-11.8.0-vfixfmc/bin/nvcc"
> 
>   is not able to compile a simple test program.
> 
>   It fails with the following output:
> 
>     Change Dir: /u/mtaylor2/software/charm/mpi-linux-x86_64-cuda/CMakeFiles/CMakeTmp
>     
>     Run Build Command(s):/usr/bin/gmake -f Makefile cmTC_4c559/fast && /usr/bin/gmake  -f CMakeFiles/cmTC_4c559.dir/build.make CMakeFiles/cmTC_4c559.dir/build
>     gmake[1]: Entering directory '/u/mtaylor2/software/charm/mpi-linux-x86_64-cuda/CMakeFiles/CMakeTmp'
>     Building CUDA object CMakeFiles/cmTC_4c559.dir/main.cu.o
>     /sw/spack/deltas11-2023-03/apps/linux-rhel8-zen3/gcc-11.4.0/cuda-11.8.0-vfixfmc/bin/nvcc      -c /u/mtaylor2/software/charm/mpi-linux-x86_64-cuda/CMakeFiles/CMakeTmp/main.cu -o CMakeFiles/cmTC_4c559.dir/main.cu.o
>     Linking CUDA executable cmTC_4c559
>     /usr/bin/cmake -E cmake_link_script CMakeFiles/cmTC_4c559.dir/link.txt --verbose=1
>     "" CMakeFiles/cmTC_4c559.dir/main.cu.o -o cmTC_4c559 
>     Error running link command: No such file or directory
>     gmake[1]: *** [CMakeFiles/cmTC_4c559.dir/build.make:99: cmTC_4c559] Error 2
>     gmake[1]: Leaving directory '/u/mtaylor2/software/charm/mpi-linux-x86_64-cuda/CMakeFiles/CMakeTmp'
>     gmake: *** [Makefile:127: cmTC_4c559/fast] Error 2

The issue: the cuda library is referenced via the variable CUDA, and the cmake script introduces an option also named CUDA. This introduces confusion when cmake tries to find cuda: enable_language(CUDA).

This PR fixes the issue by renaming the CUDA option as BUILD_CUDA. The existing cmake structure requires that there still exists an option named CUDA for the conv-mach-cuda.h discovery, so I pushed the creation of the CUDA option to after all cuda library setup.

This PR also links in the necessary cuda and HAPI libraries to charmlib.so, for use in charm4py.

[matthiasdiener]

LGTM, thanks!