Speedup time_series agg by caching current tsid ordinal, parent bucket ordinal and buck ordinal.

This avoids needlessly adding the same parent bucket ordinal or TSIDs to `BytesKeyedBucketOrds`.

Relates to elastic#74660

Author: martijnvg

modules/aggregations/src/main/java/org/elasticsearch/aggregations/bucket/timeseries/TimeSeriesAggregator.java

@@ -92,15 +92,29 @@ protected void doClose() {
     protected LeafBucketCollector getLeafCollector(AggregationExecutionContext aggCtx, LeafBucketCollector sub) throws IOException {
         return new LeafBucketCollectorBase(sub, null) {
 
+            // This helps significantly reducing time spent attempting to add bucket + tsid combos that already were added.
+            long currentTsidOrd = -1;
+            long currentBucket = -1;
+            long currentBucketOrdinal;
+
             @Override
             public void collect(int doc, long bucket) throws IOException {
+                if (currentBucket == bucket && currentTsidOrd == aggCtx.getTsidOrd()) {
+                    collectExistingBucket(sub, doc, currentBucketOrdinal);
+                    return;
+                }
+
                 long bucketOrdinal = bucketOrds.add(bucket, aggCtx.getTsid());
                 if (bucketOrdinal < 0) { // already seen
                     bucketOrdinal = -1 - bucketOrdinal;
                     collectExistingBucket(sub, doc, bucketOrdinal);
                 } else {
                     collectBucket(sub, doc, bucketOrdinal);
                 }
+
+                currentBucketOrdinal = bucketOrdinal;
+                currentTsidOrd = aggCtx.getTsidOrd();
+                currentBucket = bucket;
             }
         };
     }