Fix learning for unfolds in biabduction

Fixes that biabduction will handle unfolding in the left-hand side as as a reason to learn that the list has to be non-empty, if that was not already provable from the context.