SOLR-16989: Optimize and consolidate reuse of DocValues iterators for value retrieval

[magibney]

See: SOLR-16989

[magibney]

Proposed CHANGES.txt entry under Optimizations for 9.4.0 release:

* SOLR-16989: Optimize and consolidate reuse of DocValues iterators for value retrieval (Michael Gibney)

[magibney]

I have compared export performance on a large shard (using various sorts) against a dense field (required field id, which would not benefit at all from this optimization and could only stand to lose), and performance/latency are basically equivalent.

This stands to reason, because the overhead is basically a few int comparisons and primitive array assignments. Anyway the upshot is that for export value retrieval there should be basically no cost to introducing this in the case of dense fields, and improvements (potentially significant) to introducing this for sparse fields.

In the SolrDocumentFetcher case there should be significant improvements for all cases.

I plan to commit this early next week, pending feedback.

[dsmiley]

I don't see a code reviewer. FWIW I would have loved to reviewed this at the time; I have history in SolrDocumentFetcher.

[magibney]

I would have loved the review, sorry this didn't get on your radar at the time. Please lmk if you have any concerns about what was committed, or what could/should have been committed alongside; I certainly don't mind revisiting.