[SPARK-17346][SQL][TESTS] Fix the flaky topic deletion in KafkaSourceStressSuite

## What changes were proposed in this pull request?

A follow up Pr for SPARK-17346 to fix flaky `org.apache.spark.sql.kafka010.KafkaSourceStressSuite`.

Test log: https://amplab.cs.berkeley.edu/jenkins/job/spark-master-test-sbt-hadoop-2.4/1855/testReport/junit/org.apache.spark.sql.kafka010/KafkaSourceStressSuite/_It_is_not_a_test_/

Looks like deleting the Kafka internal topic `__consumer_offsets` is flaky. This PR just simply ignores internal topics.

## How was this patch tested?

Existing tests.

Author: Shixiong Zhu <shixiong@databricks.com>

Closes #15384 from zsxwing/SPARK-17346-flaky-test.

external/kafka-0-10-sql/src/test/scala/org/apache/spark/sql/kafka010/KafkaSourceSuite.scala

@@ -22,7 +22,6 @@ import java.util.concurrent.atomic.AtomicInteger
 import scala.util.Random
 
 import org.apache.kafka.clients.producer.RecordMetadata
-import org.scalatest.BeforeAndAfter
 import org.scalatest.time.SpanSugar._
 
 import org.apache.spark.sql.execution.streaming._
@@ -344,7 +343,7 @@ class KafkaSourceSuite extends KafkaSourceTest {
 }
 
 
-class KafkaSourceStressSuite extends KafkaSourceTest with BeforeAndAfter {
+class KafkaSourceStressSuite extends KafkaSourceTest {
 
   import testImplicits._
 
@@ -358,12 +357,6 @@ class KafkaSourceStressSuite extends KafkaSourceTest with BeforeAndAfter {
     start + Random.nextInt(start + end - 1)
   }
 
-  after {
-    for (topic <- testUtils.getAllTopicsAndPartitionSize().toMap.keys) {
-      testUtils.deleteTopic(topic)
-    }
-  }
-
   test("stress test with multiple topics and partitions")  {
     topics.foreach { topic =>
       testUtils.createTopic(topic, partitions = nextInt(1, 6))