Use ElasticSearch in single-node mode (#76)

We've been seeing datahub integration tests fail on Linux machines (e.g. datahub-project/datahub#2088). The core reason for this issue is that production deployments of ElasticSearch require some extra configuration on Linux machines.

However, we don't need a production deployment for integration tests, and as per the ES docs, we can use single-node mode to bypass these checks and make the tests succeed on Linux-based distros.

Note that this is already done in the es-7 directories

testing/elasticsearch-dao-integ-testing-docker/src/main/java/com/linkedin/metadata/testing/ElasticsearchContainerFactoryDockerImpl.java

@@ -59,7 +59,8 @@ private static TransportClient buildTransportClient(@Nonnull GenericContainerImp
   public ElasticsearchConnection start() throws Exception {
     if (_container == null) {
       _container = new GenericContainerImpl(IMAGE_NAME).withExposedPorts(HTTP_PORT, TRANSPORT_PORT)
-          .withEnv("xpack.security.enabled", "false");
+          .withEnv("xpack.security.enabled", "false")
+          .withEnv("discovery.type", "single-node");
       _container.start();
     }