Further Optimize "per partition" top-k : `ROW_NUMBER < 5` / TopK

This ticket track the follow on work from
- https://github.com/apache/datafusion/issues/6899

Where @SubhamSinghal and others have implemented specific optimizations for queries such as this (topK per partition)

```sql
SELECT ...
ROW_NUMBER() OVER (PARTITION BY value1, ORDER BY value2) as rn
WHERE
rn < 10
```

We closed https://github.com/apache/datafusion/issues/6899 to signal we have implemented the first part of the operator; This ticket tracks further optimizations

- [x] https://github.com/apache/datafusion/issues/21594
- [ ] https://github.com/apache/datafusion/issues/21596 / https://github.com/apache/datafusion/pull/23599
- [ ] https://github.com/apache/datafusion/issues/13586

The issue names no files or tests. Start by reading the prior work in issue 6899, then review unchecked issues 21596 and 13586 and pull request 23599; done means completing the remaining top-k-per-partition optimizations and validating the ROW_NUMBER query behavior.

Written by the indexing model from the issue text.