Relative Drawing Identification Complexity is Invariant to Modality in Vision-Language Models