Improve performance of aggregate expressions by implementing group accumulators

[andygrove]

What is the problem the feature request solves?

DataFusion has a fast path for aggregates that implement group accumulators.

See #1275 (comment) for one example where we saw a 4x speedup.

We should apply this change to the following Comet aggregates.

• stddev
• correlation
• covariance
• variance

Describe the potential solution

No response

Additional context

No response

[dharanad]

I would like implement group accumulators variance