MatchingDirectoryReader should not use a threaded searcher (elastic#1…

…00527)

MatchingDirectoryReader is a test reader wrapper that filters out documents
matching a particular query. For each leaf, we create an IndexSearcher,
execute the query against it and then use that as a filter for the leaf. This searcher
is created using LuceneTestCase.newSearcher() and as such may be multi-
threaded, which triggers extra index checks. For tests that are expecting certain
methods to be called against internal readers a given number of times, these extra
checks can add additional calls which then lead to a failure of test assumptions.

Because this IndexSearcher is only executed against a single leaf it will only ever use
a single thread, and so we can explicitly disable threading here.

Fixes elastic#100487
Fixes elastic#99916

test/framework/src/main/java/org/elasticsearch/index/engine/EngineTestCase.java

@@ -1513,7 +1513,7 @@ public MatchingDirectoryReader(DirectoryReader in, Query query) throws IOExcepti
                 @Override
                 public LeafReader wrap(LeafReader leaf) {
                     try {
-                        final IndexSearcher searcher = newSearcher(leaf, false);
+                        final IndexSearcher searcher = newSearcher(leaf, false, true, false);
                         searcher.setQueryCache(null);
                         final Weight weight = searcher.createWeight(query, ScoreMode.COMPLETE_NO_SCORES, 1.0f);
                         final Scorer scorer = weight.scorer(leaf.getContext());