Model Selection in Reinforcement Learning with General Function Approximations