rewritten the role to make sure it works with docker
The previous version of the code was not really working. It was impossible to use tensorflow from python or from a docker container.
It has been completely been revamped in order to follow the procedures described by Nvidia: https://docs.nvidia.com/datacenter/tesla/tesla-installation-notes/index.html#ubuntu-lts https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#post-installation-actions