Transparency Communication for Reinforcement Learning in Human-Robot Interaction