Exploring Visual Embedding Spaces Induced by Vision Transformers for Online Auto Parts Marketplaces