Run terms concurrently when cardinality is only lower than shard size

[iverase]

We currently run terms aggregations concurrently when cardinality is lower or equal to shard size. The issue is that when merging internal aggregations for cardinalities equal to the shard size, we don't know if we have truncate the list of terms, so an error is computed which is not expected.

This is happening here where the condition is size < terms.getShardSize():

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

Line 118 in ecacb2e

if (size == 0 || size < terms.getShardSize() || isKeyOrder(terms.getOrder())) {

To make sure the results are the same between concurrent and non-concurrent runs, we need to exclude situations where shard size == cardinality. This showed up in a random test.

fixes #105505

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)

[elasticsearchmachine]

Hi @iverase, I've created a changelog YAML for you.