RulePolicy: Abstract loop interface + TwoStageFallbackPolicy

[wochinge]

Proposed changes:

• fix Implement TwoStageFallbackPolicy for the RulePolicy #5915
• create abstract Loop interface which can be implemented for conversation parts which should be handled outside of ML
• adapt FormAction to loop interface
• implement TwoStageFallbackPolicy as action
• hardcode intent name of FallbackClassifier since the TwoStageFallbackClassifier relies on the intent name

Status (please check what you already did):

• added some tests for the functionality
• updated the documentation
• updated the changelog (please check changelog for instructions)
• reformat files using black (please check Readme for instructions)

[wochinge]

Quick note: the sense of the interface definition will probably be tested (and potentially changed) when doing the TwoStageFallback implementation since I'll then see whether the abstraction was correct.

[Ghostvv]

Overral looks good, there are places to think about, but if it works I think we can merge it for now