Feedback can only take one iteration

[Almendra-ab]

Garbo can receive feedback from the user and try again to look for the correct data. This only works once, after the first interaction the user is stuck with the final result and it would be very helpful to have more dialouge to continue tweaking.

There seems to have been work done to solve this but it's not working for me yet :)