When lexing input in HTML mode the lexer has to treat _all_ content of a <script> tag as plain text. This ensures that the lexer can process input such as "x <y" and "// <foo>" correctly. Fixes #70.