Re: cataloguing NOT NULL constraints

I went over the whole patch and made a very large number of additional
cleanups[1], to the point where I think this is truly ready for commit now.
There are some relatively minor things that could still be subject of
debate, such as what to name constraints that derive from PKs or from
multiple inheritance parents. I have one commented out Assert() because
of that. But other than those and a couple of not-terribly-important
XXX comments, this is as ready as it'll ever be.

I'll put it through CI soon. It's been a while since I tested using
pg_upgrade from older versions, so I'll do that too. If no problems
emerge, I intend to get this committed soon.

[1] https://github.com/alvherre/postgres/tree/catalog-notnull-9

--
Álvaro Herrera PostgreSQL Developer — https://www.EnterpriseDB.com/
"If you want to have good ideas, you must have many ideas. Most of them
will be wrong, and what you have to learn is which ones to throw away."
(Linus Pauling)