Replace slice of columns or rows return with iterator or on-demand function for memory efficiency

[DISTREAT]

Currently, when requesting columns or rows, the full result is returned as a slice, which may result in high memory use for large datasets.

Proposal:

• Replace returning a complete slice with an iterator or on-demand function to yield columns or rows progressively.
• This approach avoids loading all data at once into memory, conserving resources and improving performance when working with large CSV files.
• The new API should:
  • Be easy to adopt for clients expecting the current slice-style output.
  • Preserve or improve performance characteristics.
  • Prevent out-of-memory crashes on large or streaming inputs.

Benefits:

• Reduced peak memory usage
• Faster initial responses when using only part of the result
• Scales better for large or streaming CSV sources

Acceptance Criteria:

• Replace or supplement the current slice-returning interface with an efficient iterator-based or on-demand approach for columns or rows extraction.
• Adequate tests and documentation for new interface and migration guidance.

[DISTREAT]

I think it might be better to take on the same approach as with StructuredTable.getRow.
A custom iterator just adds unnecessary code on top.