Tokeniser should recognise long and double literals

Amazingly no one noticed that you can only write int and float literals in rules. Probably because you can use them to initialize long and double variables and Byteman will auto-coerce them.