`Fin n` for array indexes?

Happy currently has a lot of "untyped" array indexing, where nothing stops one using an index for the wrong of e.g. tokens, non-terminals, terminals.

I haven't done finite sets in Haskell for a while, so not sure the state of things these days (still `KnownNat` and friends I suppose?), but I wonder if we should try using them. It would be a lot less scary to refactor the code if these indexes wouldn't be confused for one another.

No files, tests, or entry points are named. Start by locating the untyped array-indexing sites for tokens, non-terminals, and terminals, then evaluate whether finite or typed indexes can distinguish them. Done would require an agreed approach and implementation scope, which this issue does not specify.

Written by the indexing model from the issue text.