Visuotactile-RL: Learning Multimodal Manipulation Policies with Deep Reinforcement Learning