fix(gcp_pubsub_consumer): fail health check when there are no workers

`ecpool` already returns an error even if the worker process is dead,
but we add the empty worker list clause here just for completeness.

apps/emqx_bridge_gcp_pubsub/src/emqx_bridge_gcp_pubsub_impl_consumer.erl

@@ -249,6 +249,8 @@ check_workers(InstanceId, Client) ->
             #{return_values => true}
         )
     of
+        {ok, []} ->
+            connecting;
         {ok, Values} ->
             AllOk = lists:all(fun(S) -> S =:= subscription_ok end, Values),
             case AllOk of

apps/emqx_bridge_gcp_pubsub/test/emqx_bridge_gcp_pubsub_consumer_SUITE.erl

@@ -1004,7 +1004,27 @@ t_bridge_rule_action_source(Config) ->
     ok.
 
 t_on_get_status(Config) ->
+    ResourceId = resource_id(Config),
     emqx_bridge_testlib:t_on_get_status(Config, #{failure_status => connecting}),
+    %% no workers alive
+    ?retry(
+        _Interval0 = 200,
+        _NAttempts0 = 20,
+        ?assertMatch({ok, connected}, emqx_resource_manager:health_check(ResourceId))
+    ),
+    WorkerPids = get_pull_worker_pids(Config),
+    emqx_utils:pmap(
+        fun(Pid) ->
+            Ref = monitor(process, Pid),
+            exit(Pid, kill),
+            receive
+                {'DOWN', Ref, process, Pid, killed} ->
+                    ok
+            end
+        end,
+        WorkerPids
+    ),
+    ?assertMatch({ok, connecting}, emqx_resource_manager:health_check(ResourceId)),
     ok.
 
 t_create_via_http_api(_Config) ->