Minor repair/add/improve TODOs in the Model_Intercomparison_Example notebook

[xkalet07]

Resolve minor TODOs scattered in the Model_Intercomparison_Example notebook

• Unify time series plot for all methods and include threshold line for each.
• Rebuild dictionary for plotting outputs and storing calculated metrics (or find better solution), example:

compare_methods_dict = {
      'CNN Pycomlink Torch': {
              'prediction': cml.cnn_torch_pred.values,
              'wet_dry': cml.cnn_torch_wd.values,
              'threshold': cnn_torch_threshold
              'roc': metrics.calculate_roc_curve(cmlcml.cnn_torch_pred.values, ref_wd, 0, 1),
              'confusion_matrix': skl.confusion_matrix(ref_wd, cml.cnn_torch_wd.values, labels=[1,0], normalize='true').round(decimals=4),
              'ACC': np.round(skl.accuracy_score(ref_wd, cml.cnn_torch_wd),decimals=4),
              'MCC': np.round(skl.matthews_corrcoef(ref_wd, cml.cnn_torch_wd),decimals=4),
              'F1': np.round(skl.f1_score(ref_wd, cml.cnn_torch_wd),decimals=4)
        },    
        'STD': {
              # ... same as above   
        },
        #... other methods
}

• Add ROC curve for each different cml separately, to see cml outliers and spread across the cmls.
• Move confusion matrix and other metrics into a loop for all methods.
• Show confusion matrix for all methods in one plot
• Investigate R and tl non-alignment in time domain (delays) by performing cross-correlation (other methods?).