Avoid transition to Comet in some cases if the overhead of R2C and C2R outweighs the benefit

[andygrove]

What is the problem the feature request solves?

Comet is currently slower than Spark for q16 because we transition to Comet at the end of the query just for a Sort operation and there is quite a lot of overhead of the surrounding R2C and C2R transitions (see #569 (comment) for more details).

We could implement a cost-based optimizer to avoid this. We could start off with some simple heuristics, such as "don't move to Comet just to do a sort" and then expand this into a true CBO with a cost model, although it is difficult to build an accurate cost model without statistics.

Describe the potential solution

No response

Additional context

No response