[Improvement] Trino engine will save the result set all in memory, even if the incremental collect mode selected

[iodone]

Code of Conduct

• I agree to follow this project's Code of Conduct

Search before asking

• I have searched in the issues and found no similar issues.

What would you like to be improved?

The advantage of the trino engine is the ability to stream returned data, kyuubi as a proxy should not remove this feature, storing the result set all in memory will lead to performance and OOM problems.
https://github.com/apache/incubator-kyuubi/blob/6a231519454c5d75aca2120de140c06febf7ae54/externals/kyuubi-trino-engine/src/main/scala/org/apache/kyuubi/engine/trino/TrinoStatement.scala#L81-L135

The iterator in this code is a direct conversion of ArrayBuffer, not a real stream iterator, and in each call to getNextResultSet, there is no real call to Trino.client to get the next result each time. @hddong WDYT?

How should we improve?

1. Trino's proxy implementation should be based entirely on the incremental collect model.
2. Based on IterableFetchIterator to implement Trino DataResult Iterator, we can refer to Trino-JDBC's ResultSet implementation.

Are you willing to submit PR?

• Yes I am willing to submit a PR!

[yaooqinn]

I suggest we shall always not hold data at engine side in any mode