minor: Improve equivalence handling of joins

[berkaysynnada]

Which issue does this PR close?

• Closes #.

Rationale for this change

While building the equivalence properties of joins, we can identify more precise orderings when the join is a LeftJoin and the join algorithm preserves the order of the left input (and similarly for RightJoin with the right input).

What changes are included in this PR?

In addition to the main change described above:

• Replaced hash_join_swap_subrule with the actual optimizer rule it represents, as this is the rule we should be testing
• Extended the columns in the build_sides_record_batches utility to include a new column that generates randomly ordered data. While not yet used, this column will be useful for future fuzz testing of joins.
• Increased the visibility of reorder_output_after_swap() and swap_join_projection(), as these functions are likely to be needed across optimizer rules.

Are these changes tested?

Are there any user-facing changes?

[berkaysynnada]

@comphead @xudong963 do you have any time to review this?

[comphead]

Thanks @berkaysynnada that is lgtm, however I'm not quite sure how the test asserts the changes? Should it be some performance benefit of using prefixed orderings for outer joins?

[alamb]

This PR seems stuck -- is it still waiting on some tests or is it ready to merge?

[berkaysynnada]

I just added a test showing the purpose of this change

[alamb]

🎉