Interpretable end-to-end Neurosymbolic Reinforcement Learning agents