Isometric Quotient Variational Auto-Encoders for Structure-Preserving Representation Learning In Huh