Fix AOOBE when setting min_doc_count to 0 in significant_terms (#60823)

This commit fixes the computation of the subset size on empty buckets (doc count of 0).
The aggregator test refactoring in #60683 revealed this bug.

server/src/main/java/org/elasticsearch/search/aggregations/bucket/terms/GlobalOrdinalsStringTermsAggregator.java

@@ -561,7 +561,7 @@ public class GlobalOrdinalsStringTermsAggregator extends AbstractStringTermsAggr
                 BucketUpdater<TB> updater = bucketUpdater(owningBucketOrds[ordIdx]);
                 collectionStrategy.forEach(owningBucketOrds[ordIdx], new BucketInfoConsumer() {
                     TB spare = null;
-    
+
                     @Override
                     public void accept(long globalOrd, long bucketOrd, long docCount) throws IOException {
                         otherDocCount[finalOrdIdx] += docCount;
@@ -574,7 +574,7 @@ public class GlobalOrdinalsStringTermsAggregator extends AbstractStringTermsAggr
                         }
                     }
                 });
-    
+
                 // Get the top buckets
                 topBucketsPreOrd[ordIdx] = buildBuckets(ordered.size());
                 for (int i = ordered.size() - 1; i >= 0; --i) {
@@ -797,9 +797,14 @@ public class GlobalOrdinalsStringTermsAggregator extends AbstractStringTermsAggr
             return new SignificantStringTerms.Bucket(new BytesRef(), 0, 0, 0, 0, null, format, 0);
         }
 
+        private long subsetSize(long owningBucketOrd) {
+            // if the owningBucketOrd is not in the array that means the bucket is empty so the size has to be 0
+            return owningBucketOrd < subsetSizes.size() ? subsetSizes.get(owningBucketOrd) : 0;
+        }
+
         @Override
         BucketUpdater<SignificantStringTerms.Bucket> bucketUpdater(long owningBucketOrd) throws IOException {
-            long subsetSize = subsetSizes.get(owningBucketOrd);
+            long subsetSize = subsetSize(owningBucketOrd);
             return (spare, globalOrd, bucketOrd, docCount) -> {
                 spare.bucketOrd = bucketOrd;
                 oversizedCopy(lookupGlobalOrd.apply(globalOrd), spare.termBytes);
@@ -839,7 +844,7 @@ public class GlobalOrdinalsStringTermsAggregator extends AbstractStringTermsAggr
                 bucketCountThresholds.getMinDocCount(),
                 metadata(),
                 format,
-                subsetSizes.get(owningBucketOrd),
+                subsetSize(owningBucketOrd),
                 supersetSize,
                 significanceHeuristic,
                 Arrays.asList(topBuckets)