Optimize memory allocations in query execution

[google-labs-jules]

This change optimizes the main query execution paths to reduce memory
allocations and improve performance.

• In QuerySet.Exec, the slice for scanning row values is now
  allocated only once outside the loop.
• In execIntoMaps, the slice for scannable values is also created
  only once.
• In resultIterator.Next, the slice for scan values is now part of the
  iterator struct and is reused for each row, avoiding reallocation
  on each call.

These changes significantly reduce the garbage collector pressure when
processing large result sets. I also fixed a unit test that was failing due to these changes.

[Copilot]

Pull Request Overview

This PR optimizes memory allocations in the main query execution paths to reduce garbage collector pressure when processing large result sets. The changes focus on reusing slice allocations rather than creating new ones for each row iteration.

• Pre-allocates scan value slices outside of loops in QuerySet.Exec and resultIterator
• Reuses the scanValues slice as a struct field in resultIterator to avoid per-row allocations
• Updates test setup to include the new scanValues field initialization

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
queryset.go	Pre-allocates scan values slice outside the row iteration loop and adds scanValues field to resultIterator initialization
iterator.go	Adds scanValues field to resultIterator struct and reuses it in Next() method instead of creating new slice each call
iterator_test.go	Updates test setup to initialize the new scanValues field in resultIterator

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}