Separate Spark AQE UTs for different backends

[marin-ma]

We have rewritten many AQE test cases with hard-coded Spark configurations and judging conditions in each test case, because Gluten generates shuffle output of different sizes compared to the statistics used by AQE. Currently, the CH backend and the Velox backend share the Spark UT modules, thereby sharing the AQE test cases as well. Since different backends can generate varying shuffle statistics, we may need to separate the AQE Spark UTs. This separation will allow us to modify and use different AQE configurations for each backend individually.

So far, I found 2 test suites that need to be separated:

• gluten/gluten-ut/spark[32,33,34]/src/test/scala/org/apache/spark/sql/execution/adaptive/GlutenAdaptiveQueryExecSuite.scala
• gluten/gluten-ut/spark[32,33,34]/src/test/scala/org/apache/spark/sql/execution/GlutenAdaptiveQueryExecSuite.scala

To address this issue, we can create packages org.apache.spark.sql.execution.adaptive.velox and org.apache.spark.sql.execution.adaptive.clickhouse and place the Suites for CH/Velox under these packages.

[marin-ma]

cc: @zzcclp @zhztheplayer @PHILO-HE @rui-mo