Improve index selection pass for ORDER BY clauses in KQIR

### Search before asking

- [X] I had searched in the [issues](https://github.com/apache/kvrocks/issues) and found no similar issues.

### Motivation

Currently, KQIR does not use the cost model when optimizing record sorting. Instead, it relies on specialized logic. This could be an area for improvement soon.

### Solution

_No response_

### Are you willing to submit a PR?

- [X] I'm willing to submit a PR!

Start by locating the KQIR index-selection pass for ORDER BY clauses and the existing cost model, then compare them with the specialized record-sorting logic. Define the intended selection behavior and verify that ORDER BY optimization decisions use cost estimates; the issue names no files or tests, so repository investigation is required.

Written by the indexing model from the issue text.