Pošalji tekstualnu poruku: Efficient PAC reinforcement learning in regular decision processes