multi-token lexer tests

[dherman]

Should be able to test the lexer on things that are multiple tokens, like "00.". Need a new alternative to "expected" and "error".