Sy-CON: Symmetric Contrastive Loss for Continual Self-Supervised Representation Learning