Optimize IndexedDISI by implementing binary search for sparseDocs in advance methods

### Description

If we look at below line of code, we are doing linear search if `NumDocs < 4096`.

https://github.com/apache/lucene/blob/9b68f633fdd606074e1808474f2baa23ff21df91/lucene/core/src/java/org/apache/lucene/codecs/lucene90/IndexedDISI.java#L583

This would be slow if there are repeated searches within the same Block.
I'm attaching a PR where I am trying to optimize it via BinarySearch.

Start by reading lucene/core/src/java/org/apache/lucene/codecs/lucene90/IndexedDISI.java around line 583 and trace the advance methods that search sparseDocs. Compare the current linear lookup with the surrounding IndexedDISI behavior, then verify that repeated searches preserve existing results while using binary search for sparse blocks.

Written by the indexing model from the issue text.