First feature complete version for recurrent inference machines

In this release, we have significantly expanded upon v0.0.1 and:

• The logging has been made more elaborate (better sorting in tensorboard, more descriptive console output).
• Metrics and losses are configurable and are logged separately.
• Training works for both FastMRI and the Calgary-Campinas datasets.
• Winning solution for the Calgary-Campinas challenge included.

First version

In this release code is available which trains deep learning based MRI reconstruction models on two public datasets:

• FastMRI: https://fastmri.org
• Calgary Campinas: https://sites.google.com/view/calgary-campinas-dataset/home

In this version only recurrent inference machines are supported.