Turn "data files" into actual *data* files, not code that we parse and execute at run time.

[fingolfin]

There are a bunch of issues with the current data files.

For starters, they are closely tied to use with AbstractAlgebra/Oscar. It would be kinda nice if they were in a format like JSON and in a way that would make it possible to parse this in other systems. However that's not a priority for me, my main concern is having this in OSCAR.

So from this perspective, a much more realistic goal to me seems to be using the Oscar serialization system for these files -- it is (by default) using JSON and all we need should be in it, or if not, can quickly be added. And we do strive to eventually have this format (or parts of it at least) supported in other systems, so long term it might also resolve this issue completely.

[fingolfin]

One IMHO obvious step towards this is to switch from a bunch of lists index by character types; and another bunch of lists indexed by class types; I'd rather have two lists: one whose entries are "the" character types, the other with "class types". That then only leaves the actual table, which could be a thing of its own, so for various minor reasons I'd tend to put this data into the characters, too (mainly because we will need it there in order to define additional character via tensor/linear combination/etc. -- something we don't really need for class types).