Store all associated records using a structure like:

```json
"recordLink" : [
    {
      "to" : "792361bb-4cfa-409f-9762-ab42e5a05b39",
      "origin" : "catalog",
      "created" : "bySearch",
      "title" : "Concentration en habitants dans un rayon de 500m en Wallonie - Service de visualisation REST",
      "url" : "http://localhost:8080/geonetwork/srv/api/records/792361bb-4cfa-409f-9762-ab42e5a05b39",
      "type" : "services"
    }, ...
```

in the document index. Using those information, the list of related records can be directly displayed in search result or record view.
The main drawback is that user privileges are not taken into account. The title of a record not visible by current user may be displayed.
The main advantages of this approach is that it is much faster than the related API.

A couple of issues are identified.

As we use the index to query for relations (relation not stored in children, bidirectional sibling, dataset operatedBy), the indexing must be 2 steps:
* First index all records
* Then index relations

Currently it is hard to only do a partial indexing. In this case, we only need to collect recordLink and update the doc in the index.

While editing a record, all related before and after the editing session needs to be update.
We should collect all UUIDs affected by the current session, and index them following the rule above.