Break point estimate when threshold exceeded

[gf2121]

Typically, we estimate the point value count to compare to a threshold and all we need is just a boolean which represents whether the point count is greater than this threshold. This PR proposes to parse the threshold into the intersect logic and break the recursion when the threshold is exceeded.

Dynamic pruning is a case that heavily using estimate point count so i run luceneutil for it. Here is the benchmark result on wikimedium10m:

M2 Chip

                            TaskQPS baseline      StdDevQPS my_modified_version      StdDev                Pct diff p-value
           HighTermDayOfYearSort      374.33      (1.8%)      429.74      (0.6%)   14.8% (  12% -   17%) 0.000
                      TermDTSort      388.72      (1.7%)      474.43      (1.5%)   22.1% (  18% -   25%) 0.000

Intel Chip

                            TaskQPS baseline      StdDevQPS my_modified_version      StdDev                Pct diff p-value
                      TermDTSort      123.32      (2.6%)      158.34      (2.5%)   28.4% (  22% -   34%) 0.000
           HighTermDayOfYearSort      391.59      (1.8%)      577.73      (2.7%)   47.5% (  42% -   52%) 0.000

PS: When profiling i noticed that PointTree construction cost a lot so i tried to make it reusable, this optimization also contributed to this speed-up.

[jpountz]

This is a big speedup! I left some minor comments.

[jpountz]

Nit: To make contracts of these functions clearer, I'd rather make this function private, and them have another isEstimatedPointCountGreaterThanOrEqualTo public function (and probably tagged with @lucene.internal so that we can evolve it as we want) that calls this private function?

[jpountz]

I believe that we can stop counting if cost == upperBound?

Suggested change

	} while (cost <= upperBound && pointTree.moveToSibling());
	} while (cost < upperBound && pointTree.moveToSibling());

[jpountz]

What about pulling the point tree in the constructor instead of doing it lazily (for simplicity)?

[jpountz]

LGTM

[gf2121]

Thanks for review @jpountz !

[gf2121]

Nightly benchmark:
https://home.apache.org/~mikemccand/lucenebench/TermDTSort.html
https://home.apache.org/~mikemccand/lucenebench/TermDayOfYearSort.html
https://home.apache.org/~mikemccand/lucenebench/2024.03.26.08.24.36.html

I opened a PR to add annotation:
https://github.com/mikemccand/luceneutil/pull/261/files