A Critical Investigation of Deep Reinforcement Learning for Navigation