[EPIC] Implement native columnar to row operator

[andygrove]

What is the problem the feature request solves?

Implementing a native columnar-to-row operator would reduce GC pressure by avoiding JVM allocations.

Existing PRs

• feat: Native columnar to row conversion (Phase 1) #3221
• feat: Native columnar to row conversion (Phase 2) #3266

Remaining Work

• Address review comment feat: Native columnar to row conversion (Phase 1) #3221 (comment) (Do we need to consider exportedNumRows > batchSize)
• Address review comment feat: Native columnar to row conversion (Phase 1) #3221 (comment) (Does the columnBatch need to be closed after converting?)
• Enable in plan stability tests and update golden files
• Fix Spark SQL test failures
• Add fuzz testing
• Enable feature by default

Describe the potential solution

No response

Additional context

No response