Venado cmake fixes

[dmageeLANL]

CMAKE

• Update lanl-setup.cmake for new machines.
• OpenMP
  • Using cce on venado, the OpenMP flag doesn't get set when openmp is found even though it finds OpenMP_CXX_FLAGS (which is really what's needed).
  • Removed OpenMP version check, cce's openmp (libcraymp.so) doesn't return a version on venado and it's unlikely to run into an OpenMP version over 13 years old.
  • Use OpenMP_CXX_FLAGS directly in branson_deps because on venado, using cce, cmake doesn't return anything in OpenMP::OpenMP_CXX.
• CUDA
  • Add SMS option for the compute capability of cuda samples. This is picked up from the environment. The name SMS matches the variable used by cuda-samples. Default to 80 (Ampere).
  • Add the host compiler (CXX) to the CUDA_FLAGS and set -allow-unsupported-compiler in case it's too new.
• Cray Compiler
  • The CXX_COMPILER_ID for cce is 'Clang', change the conditional to catch that.

Source

• Only output GPU status updates if ENABLE_VERBOSE_GPU_TRANSPORT is set.
• Tell if CUDA is enabled in header summary.
• Reuse total_sigma_s in transport kernel.