refactor(gcp_pubsub): always cleanup resources when calling `on_stop`

Fixes https://emqx.atlassian.net/browse/EMQX-9934

The way that this connector was structured already depended only on the resource ID.  This
PR just makes it completely independent of the resource state term, and also treats an
already absent pool as a success when stopping it.

apps/emqx_bridge_gcp_pubsub/src/emqx_bridge_gcp_pubsub_connector.erl

@@ -134,17 +134,21 @@ on_start(
     end.
 
 -spec on_stop(resource_id(), state()) -> ok | {error, term()}.
-on_stop(
-    ResourceId,
-    _State = #{jwt_config := JWTConfig}
-) ->
-    ?tp(gcp_pubsub_stop, #{resource_id => ResourceId, jwt_config => JWTConfig}),
+on_stop(ResourceId, _State) ->
+    ?tp(gcp_pubsub_stop, #{resource_id => ResourceId}),
     ?SLOG(info, #{
         msg => "stopping_gcp_pubsub_bridge",
         connector => ResourceId
     }),
-    emqx_connector_jwt:delete_jwt(?JWT_TABLE, ResourceId),
-    ehttpc_sup:stop_pool(ResourceId).
+    ok = emqx_connector_jwt:delete_jwt(?JWT_TABLE, ResourceId),
+    case ehttpc_sup:stop_pool(ResourceId) of
+        ok ->
+            ok;
+        {error, not_found} ->
+            ok;
+        Error ->
+            Error
+    end.
 
 -spec on_query(
     resource_id(),