[Core] make StringIdMap thread safe

[wumuzi520]

Why are these changes needed?

Another PR(#22817) refactoring the legacy resource data structure from ResourceSet to ResourceRequest and from SchedulingResources to Node, which depends on scheduling::ResourceID.

But, the StringIdMap is wrapped by ThreadPrivate inside BaseSchedulingID, which means the scheduling::ResourceID and scheduling::NodeID can only be visited in one single thread. However, many C++ UTs have such a scenario: restarting the GCS server multiple times in the same process, each restart will reset the GcsServer object and change a new thread, which will cause multiple threads to access StringIdMap in the different time.

Beside, the GCS may have multi-threads in the future, it is possible that in the future there will be multiple threads accessing the StringIdMap

So, this PR make StringIdMap thread safe.

Related issue number

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[scv119]

microbenchmark result:

{"single_client_get_calls_Plasma_Store": [5790.815519044334, 148.92260597977426], "single_client_put_calls_Plasma_Store": [4372.098618875549, 10.15471490167108], "multi_client_put_calls_Plasma_Store": [9563.76323784758, 57.94272163733077], "single_client_put_gigabytes": [13.943759252601867, 3.0471357661505576], "single_client_tasks_and_get_batch": [9.385313711279887, 0.04236935259512758], "multi_client_put_gigabytes": [7.10912756139562, 0.5594494022902039], "single_client_get_object_containing_10k_refs": [14.623165557495081, 0.02447091072593552], "single_client_tasks_sync": [1526.651648726323, 5.124436687275942], "single_client_tasks_async": [9206.257327530355, 65.87480980457345], "multi_client_tasks_async": [10324.538063949625, 143.88048299769758], "1_1_actor_calls_sync": [2637.0289698983574, 12.894607316293243], "1_1_actor_calls_async": [6021.686287261517, 91.8225147676578], "1_1_actor_calls_concurrent": [5018.253150418268, 39.80705381990989], "1_n_actor_calls_async": [8917.040924225807, 32.83275890958308], "n_n_actor_calls_async": [13154.73099778303, 76.71826591560907], "n_n_actor_calls_with_arg_async": [3056.814510257117, 80.7269314060778], "1_1_async_actor_calls_sync": [1767.362593685796, 9.017650518383183], "1_1_async_actor_calls_async": [3565.552973294091, 23.570889019867607], "1_1_async_actor_calls_with_args_async": [2391.7880544274262, 5.00107521063344], "1_n_async_actor_calls_async": [5997.456974780154, 42.930458124769906], "n_n_async_actor_calls_async": [8545.027368179746, 156.44482378594122], "placement_group_create/removal": [1099.906189057709, 6.568577829007929], "client__get_calls": [1518.3607888893844, 3.403318473128629], "client__put_calls": [718.9498605674409, 0.8193451515445498], "client__put_gigabytes": [0.04737091470473183, 0.00012889189344373155], "client__tasks_and_put_batch": [8973.351128895454, 66.93037447647836], "client__1_1_actor_calls_sync": [566.0864722140315, 0.9645521141581913], "client__1_1_actor_calls_async": [840.8236733302473, 7.550285579786444], "client__1_1_actor_calls_concurrent": [838.067122858635, 2.947284826728502], "client__tasks_and_get_batch": [0.7916935270820461, 0.007215860789810826], "perf_metrics": [{"perf_metric_name": "single_client_get_calls_Plasma_Store", "perf_metric_value": 5790.815519044334, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "single_client_put_calls_Plasma_Store", "perf_metric_value": 4372.098618875549, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "multi_client_put_calls_Plasma_Store", "perf_metric_value": 9563.76323784758, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "single_client_put_gigabytes", "perf_metric_value": 13.943759252601867, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "single_client_tasks_and_get_batch", "perf_metric_value": 9.385313711279887, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "multi_client_put_gigabytes", "perf_metric_value": 7.10912756139562, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "single_client_get_object_containing_10k_refs", "perf_metric_value": 14.623165557495081, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "single_client_tasks_sync", "perf_metric_value": 1526.651648726323, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "single_client_tasks_async", "perf_metric_value": 9206.257327530355, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "multi_client_tasks_async", "perf_metric_value": 10324.538063949625, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_1_actor_calls_sync", "perf_metric_value": 2637.0289698983574, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_1_actor_calls_async", "perf_metric_value": 6021.686287261517, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_1_actor_calls_concurrent", "perf_metric_value": 5018.253150418268, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_n_actor_calls_async", "perf_metric_value": 8917.040924225807, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "n_n_actor_calls_async", "perf_metric_value": 13154.73099778303, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "n_n_actor_calls_with_arg_async", "perf_metric_value": 3056.814510257117, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_1_async_actor_calls_sync", "perf_metric_value": 1767.362593685796, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_1_async_actor_calls_async", "perf_metric_value": 3565.552973294091, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_1_async_actor_calls_with_args_async", "perf_metric_value": 2391.7880544274262, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "1_n_async_actor_calls_async", "perf_metric_value": 5997.456974780154, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "n_n_async_actor_calls_async", "perf_metric_value": 8545.027368179746, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "placement_group_create/removal", "perf_metric_value": 1099.906189057709, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__get_calls", "perf_metric_value": 1518.3607888893844, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__put_calls", "perf_metric_value": 718.9498605674409, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__put_gigabytes", "perf_metric_value": 0.04737091470473183, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__tasks_and_put_batch", "perf_metric_value": 8973.351128895454, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__1_1_actor_calls_sync", "perf_metric_value": 566.0864722140315, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__1_1_actor_calls_async", "perf_metric_value": 840.8236733302473, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__1_1_actor_calls_concurrent", "perf_metric_value": 838.067122858635, "perf_metric_type": "THROUGHPUT"}, {"perf_metric_name": "client__tasks_and_get_batch", "perf_metric_value": 0.7916935270820461, "perf_metric_type": "THROUGHPUT"}]}

[jjyao]

What's C++ UT?

[scv119]

ha C++ unit tests