Updated Review — Reports Carried Forward

An update of an earlier review, which is the one place a PRISMA funnel may legitimately widen. Twenty-two reports of studies already included in the previous version enter at the included box without ever being sought this cycle — so `reports_included` exceeds `fulltext_assessed`, and the funnel-descent check would fire on a perfectly correct diagram unless the document could say why. `previous_reports` is how it says so, and the engine then checks the widened arithmetic rather than waiving it. Reports clean.