Gaze Behavior in Visual World Experiments Can be Modeled With Off-the-shelf Language-Vision Encoders