The Postgres approach to coupling locks during an ascent of the tree is
slightly different to the approach taken by Lehman and Yao.  Add a new
paragraph to the "Differences to the Lehman & Yao algorithm" section of
the nbtree README that explains the similarities and differences.