Supervised and Reinforcement Learning from Observations in Reconnaissance Blind Chess