Lucene query improvements

[bryanlb]

• Swaps Collectors to use CollectorManagers, so that we can take advantage of a custom executor (see LUCENE-10002: Deprecate IndexSearch#search(Query, Collector) in favor of IndexSearcher#search(Query, CollectorManager) - TopFieldCollectorManager & TopScoreDocCollectorManager apache/lucene#240, https://issues.apache.org/jira/browse/LUCENE-10002)
• Adjusts logic so collectors only run when data is requested instead of on every query
• Simplifies the stats collector to extend the SimpleCollector, matching the built-in collectors, https://lucene.apache.org/core/8_0_0/core/org/apache/lucene/search/SimpleCollector.html
• Removes hard-coded ramBuffer config and compound file overrides in favor of default values. This now matches the default behavior of OpenSearch merge policy.

[mansu]

A few questions about the benchmarks:

• Should the benchmarks be updated with the index sorting results?

• From the benchmarks calculating histogram takes roughly 2x the time than getting the topK results.
  0 results, 60 buckets
  Iteration 1: 38.235 ops/s
  500 results, 0 buckets
  Iteration 1: 17.951 ops/s

But calculating both histogram and topK takes less time than either:
500 results, 60 buckets
Iteration 1: 14.788 ops/s

I am curious how we can explain the speed up? Could the speed up be due to a bug in our code? Can you ensure that we have unit tests(existing or new) that test all of these outputs to make sure the outputs match the expected result?

[bryanlb]

Updated benchmarks:

New Code:
  * 500,60 - 15.263 ops/s
  * 500,0 - 68.357 ops/s
  * 0,60 - 32.471 ops/s

Master:
  * 500,60 - 14.137 op/s
  * 500,0 - 18.546 op/s
  * 0,60 - 39.498 op/s