Fix check interaction in the "serve-stale" test

Commit f351c21034 modified the
"serve-stale" system test by adding the ns3/named9.conf.in configuration
file and making the ns3 named instance load that file near the end of
the test.  However, ns3/named9.conf.in changes the
stale-answer-client-timeout setting to a very low value, which affects
all subsequent checks in tests.sh (rather than just the check that needs
the low value to be set) and may cause false positives.  Fix by
reloading configuration from ns3/named8.conf.in as soon as the check
using a very low stale-answer-client-timeout value is finished.

(cherry picked from commit cd64a3747b)

bin/tests/system/serve-stale/tests.sh

@@ -2562,6 +2562,9 @@ $RNDCCMD 10.53.0.3 serve-stale on >rndc.out.test$n.2 2>&1 || ret=1
 $DIG -p ${PORT} @10.53.0.3 a-only-slow.example AAAA >dig.out.test$n || ret=1
 grep "status: NOERROR" dig.out.test$n >/dev/null || ret=1
 grep "2001:aaaa" dig.out.test$n >/dev/null || ret=1
+# revert configuration changes introduced by this check
+copy_setports ns3/named8.conf.in ns3/named.conf
+$RNDCCMD 10.53.0.3 reload >rndc.out.test$n.1 2>&1 || ret=1
 if [ $ret != 0 ]; then echo_i "failed"; fi
 status=$((status + ret))