[SPARK-6608] [SQL] Makes DataFrame.rdd a lazy val

Before 1.3.0, `SchemaRDD.id` works as a unique identifier of each `SchemaRDD`. In 1.3.0, unlike `SchemaRDD`, `DataFrame` is no longer an RDD, and `DataFrame.rdd` is actually a function which always returns a new RDD instance. Making `DataFrame.rdd` a lazy val should bring the unique identifier back.

<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/apache/spark/5265)
<!-- Reviewable:end -->

Author: Cheng Lian <[email protected]>

Closes #5265 from liancheng/spark-6608 and squashes the following commits:

7500968 [Cheng Lian] Updates javadoc
7f37d21 [Cheng Lian] Makes DataFrame.rdd a lazy val

Author: liancheng

sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala

@@ -952,10 +952,12 @@ class DataFrame private[sql](
   /////////////////////////////////////////////////////////////////////////////
 
   /**
-   * Returns the content of the [[DataFrame]] as an [[RDD]] of [[Row]]s.
+   * Represents the content of the [[DataFrame]] as an [[RDD]] of [[Row]]s. Note that the RDD is
+   * memoized. Once called, it won't change even if you change any query planning related Spark SQL
+   * configurations (e.g. `spark.sql.shuffle.partitions`).
    * @group rdd
    */
-  def rdd: RDD[Row] = {
+  lazy val rdd: RDD[Row] = {
     // use a local variable to make sure the map closure doesn't capture the whole DataFrame
     val schema = this.schema
     queryExecution.executedPlan.execute().map(ScalaReflection.convertRowToScala(_, schema))