Add support for extracting content in a semantic way

We should try to make use of elements like `<main>`, `<article>`, `<blockquote>`, ... to figure out where an article is in the DOM. If those aren't available (or don't give the desired result), we should revert to the entire `<body>` tag.
