The buildfarm says this produces some unexpected output with
force_parallel_mode = regress.  There's probably a bug underneath
that, but for the moment just delete the test case to make the
buildfarm green again.

(I now notice that the case had also failed to get updated to follow
commit d52eaa09, which made plan_cache_mode = force_generic_plan
prevail throughout partition_prune.sql; it was thereby managing to
break a later test.  When/if we put this back in, *don't* include the
SET and RESET commands.)