[ML] Fix license feature test cleanup (elastic#84020)

Testing license features for can cause the .ml-stats index to be created
some time after the tests complete. This can cause the post-test cleanup
to fail.

This change fixes the problem by waiting for pending tasks in the first
stage cleanup, before the main cleanup methods run in the base classes.

Fixes elastic#83923

x-pack/plugin/ml/src/internalClusterTest/java/org/elasticsearch/xpack/ml/integration/TestFeatureLicenseTrackingIT.java

@@ -31,6 +31,7 @@
 import org.elasticsearch.xpack.core.ml.job.config.JobState;
 import org.elasticsearch.xpack.ml.MachineLearning;
 import org.elasticsearch.xpack.ml.MlSingleNodeTestCase;
+import org.elasticsearch.xpack.ml.support.BaseMlIntegTestCase;
 import org.junit.After;
 
 import java.time.ZonedDateTime;
@@ -57,14 +58,17 @@ public class TestFeatureLicenseTrackingIT extends MlSingleNodeTestCase {
     private final Set<String> createdPipelines = new HashSet<>();
 
     @After
-    public void cleanup() {
+    public void cleanup() throws Exception {
         for (String pipeline : createdPipelines) {
             try {
                 client().execute(DeletePipelineAction.INSTANCE, new DeletePipelineRequest(pipeline)).actionGet();
             } catch (Exception ex) {
                 logger.warn(() -> new ParameterizedMessage("error cleaning up pipeline [{}]", pipeline), ex);
             }
         }
+        // Some of the tests have async side effects. We need to wait for these to complete before continuing
+        // the cleanup, otherwise unexpected indices may get created during the cleanup process.
+        BaseMlIntegTestCase.waitForPendingTasks(client());
     }
 
     public void testFeatureTrackingAnomalyJob() throws Exception {
@@ -125,7 +129,7 @@ public void testFeatureTrackingInferenceModelPipeline() throws Exception {
             .setInferenceConfig(new ClassificationConfig(3))
             .setParsedDefinition(
                 new TrainedModelDefinition.Builder().setPreProcessors(
-                    Arrays.asList(new OneHotEncoding("other.categorical", oneHotEncoding, false))
+                    List.of(new OneHotEncoding("other.categorical", oneHotEncoding, false))
                 ).setTrainedModel(buildClassification(true))
             )
             .build();