Informed POMDP: Leveraging Additional Information in Model-Based RL