Multi-view Masked Contrastive Representation Learning for Endoscopic Video Analysis