Provide descriptives for all features, not just those in the final model

Aoife: In a few cases features of interest didn't make it into the final NNLS model. Unfortunately that also meant that they were excluded from all analyses including  feature descriptions. We were able to get feature descriptions by rerunning with Linear Regression, but would it be possible to include feature descriptives and inter-feature correlations for all features listed in the features file, not just the ones that make it into the final model? 

Jeremy: We can address this in the refactored version of RSMTool.

The feature descriptives and inter-feature correlations get calculated by the run_training_analyses() method in the Analyzer class. Currently, this method subsets the data using a selected_features parameter, but we could use something like {{all_features }}instead.

I'm assuming we would want to give users the option to include all features (whether or not they drop out of the model) or selected features in the analysis, correct? If so, we'll have to add a new configuration parameter to that effect.

Aoife: Making it an option sounds reasonable to me

Beginnen Sie mit der Methode Analyzer.run_training_analyses() und prüfen Sie, wie selected_features die Feature-Deskriptiven und die Korrelationen zwischen Features steuert. Definieren Sie den im Issue besprochenen Konfigurationsparameter und überprüfen Sie anschließend, dass Benutzer zwischen allen Features in der features-Datei und nur den ausgewählten Features wählen können und dass beide Analysen diese Auswahl widerspiegeln.

Vom Indexierungsmodell aus dem Issue-Text verfasst.