FetchOrientation.FETCH_FIRST is not handled properly

[pan3793]

Describe the bug

In SparkOperation, all order actually are treat as FETCH_NEXT

  override def getNextRowSet(order: FetchOrientation, rowSetSize: Int): TRowSet = {
    validateDefaultFetchOrientation(order)
    assertState(OperationState.FINISHED)
    setHasResultSet(true)
    val taken = iter.take(rowSetSize)
    RowSet.toTRowSet(taken.toList, resultSchema, getProtocolVersion)
  }

To Reproduce
Use HUE to connect Kyuubi server and run a query, then export the result to csv, we found the records showed in web page is lost in exported data.

Expected behavior
Hue can export all records of query result.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
Add any other context about the problem here.

[yaooqinn]

For fetch first, can we cache the result at the Kyuubi server side?

[pan3793]

Personally, I prefer to cache the result in Spark Driver side.

Considered Kybuui server will manage several sessions, if all query results cache on Kyuubi server, it will consume lots of memories, and cause high OOM risk. If Kyuubi server failover, the cache lost. If one session exit or crash, we need do some clean work to release the associated result cache.

@wangyum I remember you mentioned some optimizations on driver side query result in previous tech share but no detail description, would you like share more details on this?