feat: set stemmer to minimal

[f-necas]

Upstream PR: geonetwork/core-geonetwork#8500

---

It solves issues with elasticsearch requests.

Stemmer (as I understand) is used to optimize word computation when requests are excuted. Words with the same roots are then reduced to their root "version". So, words like "continua", "continuait" are reduced to "continu".

The lightest the stemmer is, the closest it is to a dictionnary stemmer (all words, with all declinaisons...) but it's also slower.

For a record with French word "Etablissements", some requests didn't work. It works for all languagues and default but not French.

I've set record with the same translation "Etablissements de Cleaux" for different lang and tried those ES requests :

{
  "query": {
    "bool" : {
      "should": [
        { "prefix": { "resourceTitleObject.langfre": "etabliss"}}
      ]
    }
  },
  "_source": [
    "resourceTitleObject"
  ],
  "from": 0,
  "size": 20
}

OR

{
  "query": {
    "match_phrase_prefix" : {
      "resourceTitleObject.langfre": {
         "query": "etabliss"
      }
    }
  },
  "_source": [
    "resourceTitleObject"
  ],
  "from": 0,
  "size": 20
}

[fvanderbiest]

Can you provide a bit of context in the ticket description ?
Why do we do this ? What are the expected benefits ? What are the implications ?
This will greatly improve the reviewer's experience.
Thanks :-)

[fvanderbiest]

We indeed observed issues with the datahub's gn-ui-autocomplete component which produced irrelevant results while typing "etablisse", "etablissem", "etablisseme", "etablissemen", and finallly produced relevant results with the final t on "etablissement". This can be witnessed on https://www.geo2france.fr/datahub/