End-to-End Neuro-Symbolic Reinforcement Learning with Textual Explanations