Counterfactual Explanation with Multi-Agent Reinforcement Learning for Drug Target Prediction