This simplifies the implementation of the parser for OpenType feature files, since it can now just keep token locations returned by the lexer. Before this change, the parser had to un-pack the location tuples and build new tuples that included the file path.