Add per-file and per-column change statistics to diff output

[cka-y]

Summary

Compute and include statistical summaries in the diff output, giving consumers a quick quantitative view of changes without re-processing all row data.

Notes

• Stats should reflect true counts even when row changes are truncated/capped
• Column stats only apply to the "modified" category (added/deleted rows affect all columns equally)
• Keep the computation lightweight since the data is already available during the diff pass

[cka-y]

Tasks:

• Parametrize it