tests being dependent on each other has been a code smell in QA automation for as long as I've been doing it. Why playwright offers it is strange to me