[SPARK-29611][WEBUI] Sort Kafka metadata by the number of messages

### What changes were proposed in this pull request?

Sort metadata by the number of messages in each Kafka partition

### Why are the changes needed?

help to find the data skewness problem.

### Does this PR introduce any user-facing change?

Yes, add a column count to the metadata and sort by count
![image](https://user-images.githubusercontent.com/26023240/67617886-63e06800-f81a-11e9-8718-be3a0100952e.png)

If you set `minPartitions` configurations with structure structured-streaming which doesn't have the Streaming page, my code changes in `DirectKafkaInputDStream` won't affect the WEB UI page just as it shows in the follow image

![image](https://user-images.githubusercontent.com/26023240/68020762-79520800-fcda-11e9-96cd-f0c64a36f505.png)

### How was this patch tested?

Manual test

Closes apache#26266 from dengziming/feature_ui_optimize.

Lead-authored-by: dengziming <[email protected]>
Co-authored-by: dengziming <[email protected]>
Signed-off-by: Dongjoon Hyun <[email protected]>

Author: 2 people

external/kafka-0-10/src/main/scala/org/apache/spark/streaming/kafka010/DirectKafkaInputDStream.scala

@@ -237,9 +237,10 @@ private[spark] class DirectKafkaInputDStream[K, V](
     val description = offsetRanges.filter { offsetRange =>
       // Don't display empty ranges.
       offsetRange.fromOffset != offsetRange.untilOffset
-    }.map { offsetRange =>
+    }.toSeq.sortBy(-_.count()).map { offsetRange =>
       s"topic: ${offsetRange.topic}\tpartition: ${offsetRange.partition}\t" +
-        s"offsets: ${offsetRange.fromOffset} to ${offsetRange.untilOffset}"
+      s"offsets: ${offsetRange.fromOffset} to ${offsetRange.untilOffset}\t" +
+      s"count: ${offsetRange.count()}"
     }.mkString("\n")
     // Copy offsetRanges to immutable.List to prevent from being modified by the user
     val metadata = Map(