Implement top hits aggregations

[StepanBrychta]

As part of switching from label-based filtering to ID-based filtering for concepts (see #5825), we have switched to using nested aggregations with label and ID subfields.

However, this will lead to some aggregation buckets from different ontologies having duplicate labels. Therefore, it might be helpful to surface the ontology type of each concept within each aggregation bucket returned to the frontend. The most suitable way of achieving this is probably through a top hits sub aggregation within each nested aggregation.