SelfDocSeg: A Self-Supervised vision-based Approach towards Document Segmentation