Improve performance of TPC-H q14

[andygrove]

What is the problem the feature request solves?

Initial observations:

• Query is fully native
• Similar to q19, lineitem scan takes longer with Comet
• Comet avoids an expensive C2R on 600mm rows from lineitem
• Looking at the metrics in Spark UI it looks like most operations are faster in Comet, yet the overall query wall clock time is slower
• Needs more investigation

Describe the potential solution

No response

Additional context

No response