Making (more) Impossible States Impossible

[nevrome]

I watched this short and sweet talk by Richard Feldman and was reminded of some things one should refactor in the janno data type (some of these things are still in #152): The column families Date_*, Relation_* and Contamination_* all come with states that should not be representable. Right now validation functions ensure that the input read from .janno files is correct, but it would be more reliable and more elegant to solve this on a type level.

[stschiff]

I've checked out the talk. Very nice. And indeed, that's a good idea.

[nevrome]

As discussed in #163 another helpful type in this context might be NonEmpty, which defines lists that have at least one element. We could use this e.g. for Group_Names.