Crossmodal integration enhances neural representation of task-relevant features in audiovisual face perception.