Return the max score from RandomVectorScorer.bulkScore

[mccullocht]

Return a little extra information from the bulkScore() as this will allow callers to skip results that are
below the minimum similarity needed to appear in the result set. Utilize this in hnsw and exhaustive search
to skip any topN heap manipulation for any bulk scored batches that do not meet the minimum score.

See #14013

[github-actions]

This PR has not had activity in the past 2 weeks, labeling it as stale. If the PR is waiting for review, notify the [email protected] list. Thank you for your contribution!

[benwtrent]

Yeah! This is what I had in mind. I would expect the underlying bulk scorers should be able to keep track of their max score pretty easily (without much overhead), allowing for the caller to bypass looping over uncompetitive scores

[github-actions]

This PR does not have an entry in lucene/CHANGES.txt. Consider adding one. If the PR doesn't need a changelog entry, then add the skip-changelog label to it and you will stop receiving this reminder on future updates to the PR.

[github-actions]

This PR does not have an entry in lucene/CHANGES.txt. Consider adding one. If the PR doesn't need a changelog entry, then add the skip-changelog label to it and you will stop receiving this reminder on future updates to the PR.

[mccullocht]

Integrated this into the new off heap bulk scorers.

My intuition was that I'd want to use vector instructions for this but layering it is kind of tricky since we don't want to use the vector API in any public interfaces. We could expose a VectorUtil.arrayMax() but it might be faster to use scalar instructions as we traverse the array.