Shape-centered Representation Learning for Visible-Infrared Person Re-identification