testsuite: fix occasional broker kill error

Problem: flux-start will kill all brokers after the first broker
exits.  By default the timeout to kill all the other brokers
is 20s.  A broker in a test in t3203-instance-recovery.t would
sometimes be killed due to a slow running rc3 cleanup.  This would
lead to a invalid test error.

Solution: Increase the kill timeout to 300s with the--test-exit-timeout
option.

Fixes #5290

Author: chu11

t/t3203-instance-recovery.t

@@ -9,9 +9,12 @@ test -n "$FLUX_TESTS_LOGFILE" && set -- "$@" --logfile
 
 runpty="flux ${SHARNESS_TEST_SRCDIR}/scripts/runpty.py"
 
+# N.B. Increase test-exit-timeout from default of 20s, on slower / busy
+# systems timeout can trigger and kill broker.
 test_expect_success 'start a persistent instance of size 4' '
 	mkdir -p test1 &&
-	flux start --test-size=4 -o,-Sstatedir=$(pwd)/test1 /bin/true
+	flux start --test-size=4 --test-exit-timeout=300s \
+		-o,-Sstatedir=$(pwd)/test1 /bin/true
 '
 test_expect_success 'expected broker attributes are set in recovery mode' '
 	cat >recov_attrs.exp <<-EOT &&