Updated setup.py to add and link cuda libraries

[michaelshin]

Changes made:

• Added a custom step when building the wheel for DeepView.Predict
  • This step ensures that the habitat_cuda library's rpath looks for the CUDA related libraries in the Python package directory directories and not the local directories. See screenshot below for the output of ldd
    
  • Also reverted the rpath to original path after the wheel is built in case a user is doing an editable install/development
• Added extras_requires that include the necessary CUDA libraries for each respective version of CUDA.

[michaelshin]

To install this on your machine, go to the analyzer directory and run pip install ".[cuda117]" . This will install DeepView.Predict in your Python environment.

Go to the site-packages of your python environment and go to habitat. Run ldd on the habitat_cuda.so found in that directory

[michaelshin]

• Added a build script and updated the EC2 runner yaml to use said script
• Added better tagging of the wheels so it is obvious which Python and CUDA versions were used to build it

[michaelshin]

Added an additional check on import to tell users if the runtime libraries are not linked

[codeclimate]

Code Climate has analyzed commit 3fb73c5 and detected 0 issues on this pull request.

View more on Code Climate.