Support extracting pages where there is no entities

The entities_to_xml function will crash when it is given a page with no entities at all. We should have an early return for that special case.