Byzantine-robust Federated Learning through Collaborative Malicious Gradient Filtering