[NSE-537] Increase partition number adaptively for large SHJ stages

[zhztheplayer]

What changes were proposed in this pull request?

Involve in AQE execution to tune against SHJ input partitions to avoid memory leaks.

[github-actions]

#537

[zhztheplayer]

We've changed the approach to a customizable build side input size of SHJ. To use, for example, limit build size to 500m per task:

set spark.oap.sql.columnar.shuffledhashjoin.resizeinputpartitions=true
set spark.oap.sql.columnar.shuffledhashjoin.buildsizelimit=500m

Within the settings, build side of SHJ will be split as more input partitions if the size is beyound 500m