[FEA] [SPARK-46743][SQL] Count bug after constant folding

[NVnavkumar]

apache/spark@b2f6474848f

Spark introduced a fix for Count bug handling in subqueries. Spark will prevent constant folding in subqueries that contain a COUNT-bug succeptible aggregate. There is a new configuration flag spark.sql.optimizer.decorrelateSubqueryPreventConstantHoldingForCountBug.enabled that is defaulted to true. We should ensure that our behavior is consistent with the fix that is now the default behavior.

We could fallback to GPU when the flag is set to false, but that logic is complicated since it's only COUNT in subqueries.

[mattahrens]

Priority is to investigate that we are not falling back with this new default state (true).