Fetch multiple image URLs in parallel (with a sensible concurrency limit) in `downloadImages` rather than downloading sequentially.