Learning Action Models with Conditional and Quantified Effects via Uncertainty-Guided Exploration