Provable Optimization for Adversarial Fair Self-supervised Contrastive Learning