[SPARK-26006][MLLIB] unpersist 'dataInternalRepr' in the PrefixSpan

## What changes were proposed in this pull request?
Mllib's Prefixspan - run method - cached RDD stays in cache. After run is comlpeted , rdd remain in cache.
We need to unpersist the cached RDD after run method.

## How was this patch tested?
Existing tests

Closes #23016 from shahidki31/SPARK-26006.

Authored-by: Shahid <shahidki31@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>

Author: shahidki31

mllib/src/main/scala/org/apache/spark/mllib/fpm/PrefixSpan.scala

@@ -174,6 +174,13 @@ class PrefixSpan private (
     val freqSequences = results.map { case (seq: Array[Int], count: Long) =>
       new FreqSequence(toPublicRepr(seq), count)
     }
+    // Cache the final RDD to the same storage level as input
+    if (data.getStorageLevel != StorageLevel.NONE) {
+      freqSequences.persist(data.getStorageLevel)
+      freqSequences.count()
+    }
+    dataInternalRepr.unpersist(false)
+
     new PrefixSpanModel(freqSequences)
   }