Uncertainty-Aware Reward-Free Exploration with General Function Approximation