What is the problem the feature request solves?

Comet currently tries to avoid mixing "Spark partial / Comet final" and "Comet partial / Spark final" aggregates. This is necessary in some cases due to incompatibilities between Spark and Comet for the intermediate aggregate buffer format, but in other cases this fallback is not necessary.

Describe the potential solution

Add logic to determine which aggregates are safe to have both Spark and Comet operators and do not fall back for those.

Additional context

No response