Clean up duplicate role and schema names in regression tests.

Clean up duplicate role and schema names in regression tests.

Since these names are global, using the same ones in different regression
tests creates a hazard of test failures if any two such scripts run
concurrently.  Let's establish a policy of not doing that.  In the cases
where a conflict existed, I chose to rename both sides: in principle one
script or the other could've been left in possession of the common name,
but that seems to just invite more trouble of the same sort.

There are a number of places where scripts are using names that seem
unduly generic, but in the absence of actual conflicts I left them alone.

In addition, fix insert.sql's use of "someone_else" as a role name.
That's a flat out violation of longstanding project policy, so back-patch
that change to v10 where the usage appeared.  The rest of this is just
future-proofing, as no two of these scripts are actually run concurrently
in the existing parallel_schedule.

Conflicts of schema-qualified names also exist, but will be dealt with
separately.

Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us