[ML] Do not fetch source when finding index of last state docs (#85334) (#85344)

There are two places where the ML code needs to find out which
index in the rolling sequence of state indices the previous
state document is in.

Previously the search to do this was needlessly retrieving
_source, which wastes memory.

The change tells the two searches not to fetch _source, only
the metadata.

Author: droberts195

docs/changelog/85334.yaml

@@ -0,0 +1,5 @@
+pr: 85334
+summary: Do not fetch source when finding index of last state docs
+area: Machine Learning
+type: bug
+issues: []

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/job/persistence/JobResultsPersister.java

@@ -341,6 +341,7 @@ private static SearchRequest buildQuantilesDocIdSearch(String quantilesDocId) {
         return new SearchRequest(AnomalyDetectorsIndex.jobStateIndexPattern()).allowPartialSearchResults(false)
             .source(
                 new SearchSourceBuilder().size(1)
+                    .fetchSource(false)
                     .trackTotalHits(false)
                     .query(new BoolQueryBuilder().filter(new IdsQueryBuilder().addIds(quantilesDocId)))
             );

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/process/IndexingStateProcessor.java

@@ -224,6 +224,7 @@ private String getConcreteIndexOrWriteAlias(String documentId) {
         SearchRequest searchRequest = new SearchRequest(AnomalyDetectorsIndex.jobStateIndexPattern()).allowPartialSearchResults(false)
             .source(
                 new SearchSourceBuilder().size(1)
+                    .fetchSource(false)
                     .trackTotalHits(false)
                     .query(new BoolQueryBuilder().filter(new IdsQueryBuilder().addIds(documentId)))
             );