Expose MergeRateLimiter in runOnMergeFinished hook

[zihanx]

Summary

This PR enhances the runOnMergeFinished hook in ConcurrentMergeScheduler
to provide access to OneMerge and MergeRateLimiter parameters.

Motivation

Currently, applications that subclass ConcurrentMergeScheduler to monitor
merge behavior have limited context in the runOnMergeFinished hook. The
MergeRateLimiter contains valuable metrics about merge performance
(pause time, throttling, MB/sec) that would be useful for:

• Monitoring merge performance in production
• Building custom metrics and dashboards
• Debugging merge-related issues
• Optimizing merge scheduler configurations

Changes

• Changed runOnMergeFinished from package-private to protected to allow subclassing
• Added OneMerge merge parameter for merge context
• Added MergeRateLimiter rateLimiter parameter to expose merge rate metrics

Backward Compatibility

This is a breaking change for any code that currently overrides runOnMergeFinished,
but since the method was package-private, external impact should be minimal.

Example Use Case

Applications can now override this method to emit custom metrics:

@Override
protected synchronized void runOnMergeFinished(MergeSource mergeSource, 
                                                OneMerge merge, 
                                                MergeRateLimiter rateLimiter) {
    super.runOnMergeFinished(mergeSource, merge, rateLimiter);
    
    // Emit custom metrics
    recordMetric("merge.paused.ns", rateLimiter.getTotalPausedNS());
    recordMetric("merge.stopped.ns", rateLimiter.getTotalStoppedNS());
    recordMetric("merge.mbps", rateLimiter.getMBPerSec());
}

[mikemccand]

[Disclaimer: @zihanx and I work together, at Amazon customer facing product search]

Thanks @zihanx -- these metrics are really helpful merge intensive indexing. Lucene applies backpressure (halts incoming indexing threads to give merge backlog a chance to get back under the hard merge count limit) and you are flying blind if there's no way to monitor it.

Could you add @lucene.experimental to the API javadocs to make it clear it's not necessarily a frozen API?

[mikemccand]

Looks great -- thank you @zihanx!

Oh sorry I forgot this the first time around: how about a CHANGES.txt entry describing this, and how it enables visibility on this important Lucene behavior (backpressure to indexing when there is too much merge debt)?

[zihanx]

Looks great -- thank you @zihanx!

Oh sorry I forgot this the first time around: how about a CHANGES.txt entry describing this, and how it enables visibility on this important Lucene behavior (backpressure to indexing when there is too much merge debt)?

Thanks @mikemccand! I have included an entry in CHANGES.txt and everything should be in place now :)

[mikemccand]

Thanks @zihanx -- this looks great. I'll try to merge soon if there are no objections/feeddback...