Pošalji tekstualnu poruku: On guaranteed optimal robust explanations for NLP models