Re: optimize file transfer in pg_upgrade

One other potential problem with this test is that we reuse the directory
names for each transfer mode.  That seems easy enough to fix.

FWIW, I've counted seven 006_transfer_modes failures happened during this
year. Five are from Windows animals:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=drongo&dt=2025-04-08%2004%3A18%3A15
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=fairywren&dt=2025-04-21%2008%3A03%3A06
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=fairywren&dt=2025-07-21%2012%3A35%3A58
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=fairywren&dt=2025-08-22%2000%3A04%3A05
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=drongo&dt=2025-12-28%2003%3A43%3A24

And two from culicidae, which tests EXEC_BACKEND and thus suffers from [1]
([2]):
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=culicidae&dt=2025-11-22%2012%3A31%3A23
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=culicidae&dt=2025-12-14%2018%3A24%3A48

Probably, this number could justify improving the test so that we can
identify the failure reason for sure looking at the upgrade logs. As of
now, we can only guess it, based on the animals' specifics...

[1] https://wiki.postgresql.org/wiki/Known_Buildfarm_Test_Failures#culicidae_failed_to_restart_server_due_to_incorrect_checksum_in_control_file
[2] https://www.postgresql.org/message-id/7ff9de7f-7203-cad9-76d9-45497cbedac7@gmail.com

Best regards,
Alexander