Set Default CUDA Arch

[mikemhenry]

I think we need to add something like this:

cmake_minimum_required(VERSION)

if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
  set(CMAKE_CUDA_ARCHITECTURES 75)
endif()

project(example LANGUAGES CUDA)

to our CMakeLists.txt.
That way if we are on a system that has nvcc but doesn't have a GPU, there is a fallback on which CUDA arch to build. I ran into this issue when building on conda-forge since they have nvcc in their pipeline but no GPU, so then the arch doesn't get set and the build fails.

I'll have to test this in a VM later, but let me know if you think this is unnecessary or handled in a better way.

[raimis]

The CI machines don't have GPUs, but CMake has no problems with nvcc (https://github.com/openmm/NNPOps/runs/4287486053):

-- Autodetected CUDA architecture(s): 3.5;5.0;5.2;6.0;6.1;7.0;7.5;8.0;8.6;8.6+PTX
-- Added CUDA NVCC flags for: -gencode;arch=compute_35,code=sm_35;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_52,code=sm_52;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_86,code=compute_86

[mikemhenry]

grumble I'm not sure what conda-forge is doing then with their cuda/nvcc setup since if I don't set the cuda arch I get CUDA_ARCHITECTURES is empty for target "cmTC_60975" so it must be a configuration issue then, thanks!