[core] Fix some legacy issues (#19392)

## Why are these changes needed?
There are some issues left from previous PRs.

- Put the gcs_actor_scheduler_mock_test back
- Add comment for named actor creation behavior
- Fix the comment for some flags. 

## Related issue number

BUILD.bazel

@@ -1531,20 +1531,20 @@ cc_test(
     ],
 )
 
-# cc_test(
+cc_test(
-#     name = "gcs_actor_scheduler_mock_test",
+    name = "gcs_actor_scheduler_mock_test",
-#     size = "small",
+    size = "small",
-#     srcs = [
+    srcs = [
-#         "src/ray/gcs/gcs_server/test/gcs_actor_scheduler_mock_test.cc",
+        "src/ray/gcs/gcs_server/test/gcs_actor_scheduler_mock_test.cc",
-#     ],
+    ],
-#     copts = COPTS,
+    copts = COPTS,
-#     tags = ["team:core"],
+    tags = ["team:core"],
-#     deps = [
+    deps = [
-#         ":gcs_server_lib",
+        ":gcs_server_lib",
-#         ":ray_mock",
+        ":ray_mock",
-#         "@com_google_googletest//:gtest_main",
+        "@com_google_googletest//:gtest_main",
-#     ],
+    ],
-# )
+)
 
 cc_test(
     name = "gcs_based_actor_scheduler_test",

src/ray/common/ray_config_def.h

@@ -477,7 +477,11 @@ RAY_CONFIG(int64_t, grpc_keepalive_timeout_ms, 20000);
 /// Whether to use log reporter in event framework
 RAY_CONFIG(bool, event_log_reporter_enabled, true)
 
-/// Whether to use log reporter in event framework
+/// Whether to enable register actor async.
+/// If it is false, the actor registration to GCS becomes synchronous, i.e.,
+/// core worker is blocked until GCS registers the actor and replies to it.
+/// If it is true, the actor registration is async, but actor handles cannot
+/// be passed to other worker until it is registered to GCS.
 RAY_CONFIG(bool, actor_register_async, true)
 
 /// Event severity threshold value

src/ray/core_worker/core_worker.cc

@@ -1800,6 +1800,9 @@ Status CoreWorker::CreateActor(const RayFunction &function,
           },
           "ActorCreator.AsyncRegisterActor");
     } else {
+      // For named actor, we still go through the sync way because for
+      // functions like list actors these actors need to be there, especially
+      // for local driver. But the current code all go through the gcs right now.
       auto status = actor_creator_->RegisterActor(task_spec);
       if (!status.ok()) {
         return status;

src/ray/gcs/gcs_server/test/gcs_actor_scheduler_mock_test.cc

@@ -84,7 +84,7 @@ TEST_F(GcsActorSchedulerTest, KillWorkerLeak1) {
   actor_data.set_actor_id(actor_id.Binary());
   auto actor = std::make_shared<GcsActor>(actor_data);
   std::function<void(const Status &, const rpc::RequestWorkerLeaseReply &)> cb;
-  EXPECT_CALL(*raylet_client, RequestWorkerLease(Matcher<const rpc::TaskSpec &>(), _, _))
+  EXPECT_CALL(*raylet_client, RequestWorkerLease(An<const rpc::TaskSpec &>(), _, _))
       .WillOnce(testing::SaveArg<1>(&cb));
   // Ensure actor is killed
   EXPECT_CALL(*core_worker_client, KillActor(_, _));
@@ -113,7 +113,7 @@ TEST_F(GcsActorSchedulerTest, KillWorkerLeak2) {
   rpc::ClientCallback<rpc::RequestWorkerLeaseReply> request_worker_lease_cb;
   // Ensure actor is killed
   EXPECT_CALL(*core_worker_client, KillActor(_, _));
-  EXPECT_CALL(*raylet_client, RequestWorkerLease(Matcher<const rpc::TaskSpec &>(), _, _))
+  EXPECT_CALL(*raylet_client, RequestWorkerLease(An<const rpc::TaskSpec &>(), _, _))
       .WillOnce(testing::SaveArg<1>(&request_worker_lease_cb));
 
   std::function<void(ray::Status)> async_put_with_index_cb;