pg_dump/pg_restore is very slow. It might take days if upgrading relatively
small cluster of 200Gb. (I got one from persoinal using of matrix-synapse for
several years, so even a single person can have cluster that big now)

My suggestion is either completely make `-m upgrade` a new default, or add new
method `auto` that will use either `dump` of `upgrade` depending on cluster
size.

pg_upgrade is safe and fast, and is widely used in production, and I see no
reason for not using it all the time by default.

And since `pg_upgradecluster` is started while upgrading the system, I would
also suggest to add a menu that asks for upgrade method. May be

And since I both perl developer and deal with PostgreSQL core code on regular
basis I can provide patches for these functionality, if Debian PostgreSQL
Maintainers approve the idea.