Native Shuffle failed on SF3000 TPCH Q10 w/ 25K batch size

[zhouyuan]

Describe the bug
the same code runs OK on 20k batch size, so it looks like some wrong data structure used

Job aborted due to stage failure: Task 250 in stage 573.0 failed 4 times, most recent failure: Lost task 250.3 in stage 573.0 (TID 215824) (sr260 executor 71): java.lang.RuntimeException: Native split: splitter split failed - Error during calling Java code from native code
	at com.intel.oap.vectorized.ShuffleSplitterJniWrapper.split(Native Method)
	at org.apache.spark.shuffle.ColumnarShuffleWriter.write(ColumnarShuffleWriter.scala:162)
	at org.apache.spark.shuffle.ShuffleWriteProcessor.write(ShuffleWriteProcessor.scala:59)
	at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:99)

To Reproduce
Run SF3000 TPCH Q10 w/ 25k batch size

Expected behavior
run ok w/ 25k or bigger batch size

Additional context
N/A

[zhouyuan]

@PHILO-HE

[zhouyuan]

close by #594