Fix handling of empty keyword in terms aggregation (#34457)

Empty values on keyword fields are filtered by the `map` execution mode
of the `terms` aggregation. This commit restores them as valid buckets.

Closes #34434

Author: jimczi

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

@@ -93,7 +93,7 @@ public void collect(int doc, long bucket) throws IOException {
                         if (includeExclude != null && !includeExclude.accept(bytes)) {
                             continue;
                         }
-                        if (previous.get().equals(bytes)) {
+                        if (i > 0 && previous.get().equals(bytes)) {
                             continue;
                         }
                         long bucketOrdinal = bucketOrds.add(bytes);

server/src/test/java/org/elasticsearch/search/aggregations/bucket/terms/TermsAggregatorTests.java

@@ -158,13 +158,17 @@ public void testSimple() throws Exception {
                 document.add(new SortedSetDocValuesField("string", new BytesRef("b")));
                 indexWriter.addDocument(document);
                 document = new Document();
+                document.add(new SortedSetDocValuesField("string", new BytesRef("")));
                 document.add(new SortedSetDocValuesField("string", new BytesRef("c")));
                 document.add(new SortedSetDocValuesField("string", new BytesRef("a")));
                 indexWriter.addDocument(document);
                 document = new Document();
                 document.add(new SortedSetDocValuesField("string", new BytesRef("b")));
                 document.add(new SortedSetDocValuesField("string", new BytesRef("d")));
                 indexWriter.addDocument(document);
+                document = new Document();
+                document.add(new SortedSetDocValuesField("string", new BytesRef("")));
+                indexWriter.addDocument(document);
                 try (IndexReader indexReader = maybeWrapReaderEs(indexWriter.getReader())) {
                     IndexSearcher indexSearcher = newIndexSearcher(indexReader);
                     for (TermsAggregatorFactory.ExecutionMode executionMode : TermsAggregatorFactory.ExecutionMode.values()) {
@@ -181,15 +185,17 @@ public void testSimple() throws Exception {
                         indexSearcher.search(new MatchAllDocsQuery(), aggregator);
                         aggregator.postCollection();
                         Terms result = (Terms) aggregator.buildAggregation(0L);
-                        assertEquals(4, result.getBuckets().size());
-                        assertEquals("a", result.getBuckets().get(0).getKeyAsString());
+                        assertEquals(5, result.getBuckets().size());
+                        assertEquals("", result.getBuckets().get(0).getKeyAsString());
                         assertEquals(2L, result.getBuckets().get(0).getDocCount());
-                        assertEquals("b", result.getBuckets().get(1).getKeyAsString());
+                        assertEquals("a", result.getBuckets().get(1).getKeyAsString());
                         assertEquals(2L, result.getBuckets().get(1).getDocCount());
-                        assertEquals("c", result.getBuckets().get(2).getKeyAsString());
-                        assertEquals(1L, result.getBuckets().get(2).getDocCount());
-                        assertEquals("d", result.getBuckets().get(3).getKeyAsString());
+                        assertEquals("b", result.getBuckets().get(2).getKeyAsString());
+                        assertEquals(2L, result.getBuckets().get(2).getDocCount());
+                        assertEquals("c", result.getBuckets().get(3).getKeyAsString());
                         assertEquals(1L, result.getBuckets().get(3).getDocCount());
+                        assertEquals("d", result.getBuckets().get(4).getKeyAsString());
+                        assertEquals(1L, result.getBuckets().get(4).getDocCount());
                     }
                 }
             }