Now that the ordering of DROP messages ought to be stable everywhere,
we should not need these kluges of hiding DETAIL output just to avoid
unstable ordering.  Hiding it's not great for test coverage, so
let's undo that where possible.

In a small number of places, it's necessary to leave it in, for
example because the output might include a variable pg_temp_nnn
schema name.  I also left things alone in places where the details
would depend on other regression test scripts, e.g. plpython_drop.sql.

Perhaps buildfarm experience will show this to be a bad idea,
but if so I'd like to know why.

Discussion: https://postgr.es/m/E1h6eep-0001Mw-Vd@gemulon.postgresql.org