Quantifying and Mitigating Privacy Risks of Contrastive Learning