[Lexer] Fix offset capture for utf strings

[sirian]

at this moment PREG_SPLIT_OFFSET_CAPTURE doesn't maintain utf-8 strings. I want to have multibyte offset