[SPARK-4042][SQL] Append columns ids and names before broadcast

Append columns ids and names before broadcast ```hiveExtraConf```  in ```HadoopTableReader```.

Author: scwf <wangfei1@huawei.com>

Closes #2885 from scwf/HadoopTableReader and squashes the following commits:

a8c498c [scwf] append columns ids and names before broadcast

sql/hive/src/main/scala/org/apache/spark/sql/hive/execution/HiveTableScan.scala

@@ -68,6 +68,9 @@ case class HiveTableScan(
   @transient
   private[this] val hiveExtraConf = new HiveConf(context.hiveconf)
 
+  // append columns ids and names before broadcast
+  addColumnMetadataToConf(hiveExtraConf)
+
   @transient
   private[this] val hadoopReader = 
     new HadoopTableReader(attributes, relation, context, hiveExtraConf)
@@ -105,8 +108,6 @@ case class HiveTableScan(
     hiveConf.set(serdeConstants.LIST_COLUMNS, relation.attributes.map(_.name).mkString(","))
   }
 
-  addColumnMetadataToConf(hiveExtraConf)
-
   /**
    * Prunes partitions not involve the query plan.
    *