Omit maxScoreCollector for field collapsing when sort by score descending

[gaobinlong]

Description

Similar to SimpleTopDocsCollectorContext, we can omit the MaxScoreCollector when sorting by score in CollapsingTopDocsCollectorContext, this can reduce the search latency and cpu usage.

By a benchmark test against the big5 index and other indices, this change can achieve 10%-20% performance improvement in both concurrent segment search enabled and disabled.

The query DSL on big5 index is:

GET big5/_search
{
  "query": {
    "match": {
      "message": "50-136-239-27"
    }
  },
  "collapse": {
    "field": "process.name.keyword"
  }
}

, the benchmark result:

Before:

|                                        50th percentile latency | collapsing |     740.061 |     ms |
|                                        90th percentile latency | collapsing |     748.504 |     ms |
|                                        99th percentile latency | collapsing |     771.312 |     ms |
|                                      99.9th percentile latency | collapsing |     843.527 |     ms |
|                                       100th percentile latency | collapsing |     857.434 |     ms |
|                                   50th percentile service time | collapsing |     738.741 |     ms |
|                                   90th percentile service time | collapsing |     747.404 |     ms |
|                                   99th percentile service time | collapsing |     770.124 |     ms |
|                                 99.9th percentile service time | collapsing |     842.116 |     ms |
|                                  100th percentile service time | collapsing |     856.587 |     ms |


After

|                                        50th percentile latency | collapsing |     568.893 |     ms |
|                                        90th percentile latency | collapsing |     579.025 |     ms |
|                                        99th percentile latency | collapsing |      621.08 |     ms |
|                                      99.9th percentile latency | collapsing |     660.756 |     ms |
|                                       100th percentile latency | collapsing |     709.702 |     ms |
|                                   50th percentile service time | collapsing |     567.327 |     ms |
|                                   90th percentile service time | collapsing |     577.377 |     ms |
|                                   99th percentile service time | collapsing |      619.84 |     ms |
|                                 99.9th percentile service time | collapsing |      659.55 |     ms |
|                                  100th percentile service time | collapsing |     708.664 |     ms |

In addition, this PR fixes the bug that max_score is null when we have multiple sort fields and _score is used as a primary sort.

Related Issues

Resolves #19179 and partly resolves #19180.

Check List

• Functionality includes testing.
  - [ ] API changes companion pull request created, if applicable.
  - [ ] Public documentation issue/PR created, if applicable.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[github-actions]

❌ Gradle check result for 23018ff: FAILURE

Please examine the workflow log, locate, and copy-paste the failure(s) below, then iterate to green. Is the failure a flaky test unrelated to your change?

[github-actions]

❕ Gradle check result for 23018ff: UNSTABLE

Please review all flaky tests that succeeded after retry and create an issue if one does not already exist to track the flaky failure.

[codecov]

Codecov Report

❌ Patch coverage is 85.00000% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 72.80%. Comparing base (2eb9b53) to head (23018ff).

Files with missing lines	Patch %	Lines
...ensearch/search/query/TopDocsCollectorContext.java	85.00%	1 Missing and 2 partials ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##               main   #19181      +/-   ##
============================================
- Coverage     72.87%   72.80%   -0.08%     
+ Complexity    69645    69608      -37     
============================================
  Files          5656     5656              
  Lines        319999   320010      +11     
  Branches      46335    46338       +3     
============================================
- Hits         233191   232968     -223     
- Misses        67907    68134     +227     
- Partials      18901    18908       +7     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[kkewwei]

I’m curious why we don’t retrieve it from the maxScoreCollector (as shown in line 271), and instead need to get the maximum score from topDocs.scoreDocs[0]?