Investigating the Impact of Observation Space Design Choices On Training Reinforcement Learning Solutions for Spacecraft Problems