improve parsing speed

The lexer speed (reading a string from memory) is about 2,5 millions tokens / sec (depends on CPU).

It's about 3 times more slower than original PHP algorithm, but about 16 times more rapid than a jison based version (which is also more rapid than a regex version ...)

The lexer algorithm could be improved a bit, but there will not be much gain here.

The parser sometimes is 10x slower than the lexer, there is a lot of room to improve the speed, the most cost effective part is recursive parsing on structures like arrays, property chains or string concatenations.