Fixing flaky test -

The `SchedulerPoolTest` requires a separate scheduler configuration file
which need to be set at the time of starting the SparkContext. Hence,
stopping the SparkContext (if started by previously run suite) before
running this test to force create new SparkContext with configuration
set with correct scheduler pool config file.

cluster/src/test/scala/org/apache/spark/sql/SchedularPoolTest.scala

@@ -39,6 +39,7 @@ class SchedulerPoolTest extends SnappyFunSuite with BeforeAndAfter with BeforeAn
 
   override def beforeAll() {
     super.beforeAll()
+    sc.stop()
     kafkaTestUtils = new KafkaTestUtils
     kafkaTestUtils.setup()
   }