Disable buffering to disk for Azure output streams

apache/hadoop@acffe20 introduced logic to ABFS Output streams buffering blocks to disk (by default).
This was introduced in Trino during Hadoop upgrade trinodb@343b908. We want to disable this and stick with the legacy behavior of keeping the blocks in memory

lib/trino-hdfs/src/main/java/io/trino/hdfs/azure/TrinoAzureConfigurationInitializer.java

@@ -27,6 +27,8 @@
 
 import static com.google.common.base.Preconditions.checkArgument;
 import static java.lang.String.format;
+import static org.apache.hadoop.fs.azurebfs.constants.ConfigurationKeys.DATA_BLOCKS_BUFFER;
+import static org.apache.hadoop.fs.store.DataBlocks.DATA_BLOCKS_BUFFER_ARRAY;
 
 public class TrinoAzureConfigurationInitializer
         implements ConfigurationInitializer
@@ -117,6 +119,9 @@ public void initializeConfiguration(Configuration config)
 
         // do not rely on information returned from local system about users and groups
         config.set("fs.azure.skipUserGroupMetadataDuringInitialization", "true");
+
+        // disable buffering Azure output streams to disk(default is DATA_BLOCKS_BUFFER_DISK)
+        config.set(DATA_BLOCKS_BUFFER, DATA_BLOCKS_BUFFER_ARRAY);
     }
 
     private static Optional<String> dropEmpty(Optional<String> optional)