chore: Don't transform the HashAggregate to CometHashAggregate if Comet shuffle is disabled

[viirya]

Which issue does this PR close?

Closes #.

Rationale for this change

What changes are included in this PR?

How are these changes tested?

[mbutrovich]

Do the benchmarks in

datafusion-comet/spark/src/test/scala/org/apache/spark/sql/benchmark/CometAggregateBenchmark.scala

Line 4 in 84cccf7

* distributed with this work for additional information

need to be updated? Part of my exploration today (that motivates this PR) is that the CometBaseBenchmark and by extension the CometAggregateBenchmark does not set Comet at the shuffle manager. CometExecBenchmark does set the shuffle manager:

datafusion-comet/spark/src/test/scala/org/apache/spark/sql/benchmark/CometExecBenchmark.scala

Line 44 in 84cccf7

"spark.shuffle.manager",

[viirya]

Yea, we should update CometAggregateBenchmarks to enable shuffle for it.

[mbutrovich]

This should help a lot for #987! Thanks for the quick fix!

[viirya]

Thanks @kazuyukitanimura @mbutrovich