LUCENE-9298: Improve RAM accounting in BufferedUpdates when deleted doc IDs and terms are cleared

[bringyou]

the method clearDeletedDocIds in BufferedUpdates.java has a bug, it can't reset bytesUsed correctly.

void clearDeletedDocIds() {
  deleteDocIDs.clear();
  bytesUsed.addAndGet(-deleteDocIDs.size() * BufferedUpdates.BYTES_PER_DEL_DOCID);
}

this PR will fix it.

[bringyou]

@s1monw sorry to bother, but can you take a look?

[dnhatn]

Change looks good to me. Would you mind adding a small test for this issue? Thanks @bringyou!

[bringyou]

Change looks good to me. Would you mind adding a small test for this issue? Thanks @bringyou!

sorry for the delay~ add a test for BufferedUpdates and change a bit more code, please take another look @dnhatn

[s1monw]

the test looks good. I like the change, I left one comment

[s1monw]

looks great I left one suggestions that we should merge in. Was my fault.

[s1monw]

I ran tests and we are ready to go. Please add a changes entry like this:

diff --git a/lucene/CHANGES.txt b/lucene/CHANGES.txt
index b5bc71cb759..2daa8f3c71f 100644
--- a/lucene/CHANGES.txt
+++ b/lucene/CHANGES.txt
@@ -238,6 +238,9 @@ Improvements
 * LUCENE-9171: QueryBuilder can now use BoostAttributes on input token streams to selectively
   boost particular terms or synonyms in parsed queries. (Alessandro Benedetti, Alan Woodward)
 
+* LUCENE-9298: Improve RAM accounting in BufferedUpdates when deleted doc IDs and terms are cleared.
+  (Your Name, Simon Willnauer)
+
 Optimizations
 ---------------------

[bringyou]

@s1monw thanks for the reviews, I add the changelogs and edit issue's title

[s1monw]

thanks!!