pgsql: Close yet another race condition in replication slot test code

Close yet another race condition in replication slot test code

Buildfarm shows that this test has a further failure mode when a
checkpoint starts earlier than expected, so we detect a "checkpoint
completed" line that's not the one we want.  Change the config to try
and prevent this.

Per buildfarm

Author: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Discussion: https://postgr.es/m/20210729.162038.534808353849568395.horikyota.ntt@gmail.com

Branch
------
REL_13_STABLE

Details
-------
https://git.postgresql.org/pg/commitdiff/41d27ee7b870c1a1213704d3c020a01eb55799b0

Modified Files
--------------
src/test/recovery/t/019_replslot_limit.pl | 14 ++++++++++++--
1 file changed, 12 insertions(+), 2 deletions(-)