cs_walrep_2: fix races in test_negative

pg_ctl and gpstart appear to be running simultaneously, because we're
not waiting for pg_ctl to finish before moving to the next call.
(subprocess.Popen() doesn't return the process's return code -- it
returns a Popen instance.)

cs_walrep_2 in the master pipeline is currently failing roughly 50% of
the time due to a crash during test_negative. Let's see if this helps at
all.

src/test/tinc/tincrepo/mpp/gpdb/tests/storage/walrepl/walreceiver/test_negative.py

@@ -123,7 +123,7 @@ class neg_test(StandbyRunMixin, MPPTestCase):
         os.remove(os.path.join(orig_master.datadir,'recovery.conf'))
 
         logger.info('Stop the original master again...')
-        rc = subprocess.Popen('pg_ctl stop -D ' + orig_master.datadir + ' -m immediate',
+        subprocess.check_call('pg_ctl stop -D ' + orig_master.datadir + ' -m immediate',
                               shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
 
         # Perform gpstart to get the original master (& cluster) back again