Tokenize purely using regexes

[mvorisek]

Match next token using single regex only, +18% speedup.

In the future, Tokenizer::makeTokenizeRegexes() method can be made public to allow grammar override for each token type.