Grounding should handle partially interpreted functions better

[Broes De Cat]

For example, only write out the possibly trues and not also all possibly falses. If there is only one possibly true, it should be directly replaced during translation.

In the provided example, it goes wrong because no clean is done, but that only indicates the deeper lying issue.

[Broes De Cat]

Solution: when adding the fd variable to the grounding, the translator should enumerate all still-allowed assignments as the domain of the variable. additionally, whenever translating a term, if only one value is still possible, it should return that value.