Self-supervised visual place recognition for colonoscopy sequences