By checking maxTokens inside the createToken function we can be sure that it's applied consistently.

I've also moved ParserOptions.lexer to be an internal property (it should only be used by the schema coordinates parser, it was never intended to be a public interface) and implemented an assertion that forbids lexer and maxTokens together (since if they were both set, maxTokens would be ignored). I've also marked the SchemaCoordinateLexer as internal; we don't expose it via index.ts anyway, but best to signal the intent anyway.

I'm not super happy about mutating the internal state of the lexer from createToken - really I'd rather createToken were a method on the lexer class - but I don't want the two Lexers to have to implement it each and we don't like to extend classes so I think we can accept it in this one place. I'm not happy about it though, so if you have ideas for a better approach, please go ahead!