Build updates and improvements

[romerojosh]

This PR makes a few updates to the CMake files and docker build scripts based on recent issues/suggestions:

• Update Docker build to NVHPC SDK 23.7: This is related to CMakeList.txt doesn't support CUDA Arch 89 #8 as cc89 support was introduced in the 23.7 version of the compilers. Updating to this version so that the docker build will support this use case.

• Build fixes from cmake cannot locate MPI fortran (from NVHPC 23.7) #3:

  • Explicitly setting CMake CMP0057 policy
  • Use g++ for compiling CXX files to support __rdtsc() usage in torch

• Fixes docker build uses large amount of memory when running with more than 4 cores #9:

  • Use -j$(nproc) in Dockerfile make commands to limit parallel build processes to number of cores. Can be further limited by enforcing CPU core limits in docker build command if neccessary.