- Update the selectors for breadcrumb navigation (for section title),
  and stop using lastChild which can be a text node.
- Use page URL directly as the URL field, because URL in JSON data may
  point to a different domain.
- Prefer title scraped from page body, because the one from JSON data
  may fail to match the former, possibly due to social media SEO.
- More robust determination of authors, accounting for different data
  formats across different sites.

See also:
https://forums.zotero.org/discussion/105950/attempt-to-save-using-embedded-metadata