Fix issue in recovery test 041_checkpoint_at_promote

The phase of the test waiting for a restartpoint to complete was not
working as intended, due to a log_contains() call incorrectly
written.

The problem reported by the author could be simply reproduced by
removing the injection_points_wakeup() call: the test succeeds rather
than waiting for the restartpoint completion.  In most cases, the
restartpoint completion is fast enough that the test offered the wanted
coverage.  On slow machines, it could have become unreliable.

Oversight in 6782709df8.

Author: Nitin Jadhav
Discussion: https://postgr.es/m/CAMm1aWa_6u+o52r7h7G6pX-oWD0Qraf0ee17Ma50qxGS0B_Rzg@mail.gmail.com
Backpatch-through: 17

src/test/recovery/t/041_checkpoint_at_promote.pl

@@ -110,7 +110,7 @@ $node_standby->safe_psql('postgres',
 my $checkpoint_complete = 0;
 foreach my $i (0 .. 10 * $PostgreSQL::Test::Utils::timeout_default)
 {
-	if ($node_standby->log_contains("restartpoint complete"), $logstart)
+	if ($node_standby->log_contains("restartpoint complete", $logstart))
 	{
 		$checkpoint_complete = 1;
 		last;