Rethinking the Representational Continuity: Towards Unsupervised Continual Learning