increased precision of metrics in csv files, added code for ranking

[FabianIsensee]

Ranking code

increased precision of metrics in csv files to prevent ties as much as possible, added ranking code

see ranking.py

Also added nnUNet_summary.csv to highlight how the ranking works. Feel free to ask questions :-)

Checklist

• [x] Merged latest master
• [NOPE] Updated version number in README.md
• [NOPE ] Added changes to changelog.md
• [NOPE ] Updated version number in setup.py
• [NOPE] (only when updating dataset) Ran annotation.import to completion
• [NOPE] (only when updating dataset) Updated Surface Dice tolerances in labels.py (execute sample_segmentations.py, and after that compute_tolerances.py. Then put the new values in. Do not re-use sampled segmentations from prior dataset versions!)

If you want me to increase the version number please let me know

Best, Fabian

[FabianIsensee]

I also updated the docker instructions and added a dockerfile + inference script for the nnunet cascade