[SPARK-21623][ML] fix RF doc

## What changes were proposed in this pull request?

comments of parentStats in RF are wrong.
parentStats is not only used for the first iteration, it is used with all the iteration for unordered features.

## How was this patch tested?

Author: Peng Meng <peng.meng@intel.com>

Closes #18832 from mpjlu/fixRFDoc.

mllib/src/main/scala/org/apache/spark/ml/tree/impl/DTStatsAggregator.scala

@@ -78,9 +78,9 @@ private[spark] class DTStatsAggregator(
 
   /**
    * Array of parent node sufficient stats.
-   *
-   * Note: this is necessary because stats for the parent node are not available
-   *       on the first iteration of tree learning.
+   * Note: parent stats need to be explicitly tracked in the [[DTStatsAggregator]] for unordered
+   *       categorical features, because the parent [[Node]] object does not have [[ImpurityStats]]
+   *       on the first iteration.
    */
   private val parentStats: Array[Double] = new Array[Double](statsSize)