metrics.py, models and utils/plot_info.py were changed to work with more than one class (multiclass).
ToDo:
	- fix metrics to work with multiclass (Konrad)
	- make number of classes flexible (based on mask) (Aleks)

The user can now see the percentage of specific classes (features) which appear in the data.
ToDo: Right now hard coded class names in prepare_data.py. Next step: read feature labels from config.

Added config option to prepare_data.py to get the labels from config.yml.
Changed variable names from classes and labels to "features" to make it (hopefully) less confusing.
Added two asserts regarding problems that might appear.

- Added and already tested three warnings.
- config works now with a dictionary instead of array.
	example: data_classes: {0 : "background", 1 : "slope", 2 : "lake", 3 : "class3", 7 : "class7", 8 : "class8"}