Parallel Append subplan order instability on aye-aye

Hi,

Here's a one-off regression test failure of a sort that commit
624e440a intended to fix.  a_star unexpectedly sorted higher.  I
checked the space weather forecast for this morning but no sign of
solar flares.  More seriously, it did the same in all 3 Parallel
Append queries.  Recent commits look irrelevant.  Could a UDP stats
packet dropped on the floor cause that?  Otherwise maybe you'd need a
weird result from FileSize() to explain it.  Based on log output no
other tests ran around the same time.

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=aye-aye&dt=2019-05-19%2018%3A30%3A10

-- 
Thomas Munro
https://enterprisedb.com