Better handle token datatype changes

[luqmana]

Every token is one of a few types based on its data size: Bool, Byte, Word, Dword. We currently collapse all the defined tokens into a single flat namespace. This does make it a bit annoying if a token gets moved from one to another (e.g. Byte -> Word to accommodate more values in a later version).

One option is to update the codegen to get rid of the flat namespace (#120 (comment)). The json and binary formats already disambiguate the differently sized token entries.