Current Lexer has a function to skip UTF-8 BOM. But its behavior has some bugs. It skips shebang at the beginning of file, but does not with BOM. A token should be prefix operator at the beginning...