Main steps

It doesn't do what it says in nnUNet_find_best_configuration -h, so be aware! So here are some main steps of what it does:

1. Determine the final postprocessing after all five folds are completed.
2. Get the mean dice of each model declared in argument -m
3. Get the mean dice of every combination of 2 models
4. Determine the best model of ensemble of 2 models
5. Display and save command lines to run the inference in RESULT_FOLDER/ENSEMBLE_NAME/prediction_commands.txt
6. Save mean dice of all models and ensembles in RESULT_FOLDER/ENSEMBLE_NAME/summary.csv