This removes the yodl dependency, which is sometimes hard to track down.
Instead, this uses a python3 script that leverages the cmarkgfm library
to turn the source file into html. Then, the script parses the html in
order to turn the tag stream into a nroff stream using a simple state
machine. While it's doing that it also implements one added format rule
that turns an ordinal list that starts at 0 into a description list
(since markdown doesn't have an easy description list idiom).