[SPARK-25266][CORE] Fix memory leak in Barrier Execution Mode

## What changes were proposed in this pull request?

BarrierCoordinator uses Timer and TimerTask. `TimerTask#cancel()` is invoked in ContextBarrierState#cancelTimerTask but `Timer#purge()` is never invoked.

Once a TimerTask is scheduled, the reference to it is not released until `Timer#purge()` is invoked even though `TimerTask#cancel()` is invoked.

## How was this patch tested?

I checked the number of instances related to the TimerTask using jmap.

Closes apache#22258 from sarutak/fix-barrierexec-oom.

Authored-by: sarutak <sarutak@oss.nttdata.co.jp>
Signed-off-by: Xiangrui Meng <meng@databricks.com>

core/src/main/scala/org/apache/spark/BarrierCoordinator.scala

@@ -123,6 +123,7 @@ private[spark] class BarrierCoordinator(
     private def cancelTimerTask(): Unit = {
       if (timerTask != null) {
         timerTask.cancel()
+        timer.purge()
         timerTask = null
       }
     }

core/src/main/scala/org/apache/spark/BarrierTaskContext.scala

@@ -140,6 +140,7 @@ class BarrierTaskContext(
         throw e
     } finally {
       timerTask.cancel()
+      timer.purge()
     }
   }