Feature/9450 improve search relevance calculation

[shaunahu]

Before this change, a search word was scored twice: once on title or description, and again on the synonyms field. That happened even when the word was not an acronym, so ordinary words and full names like "ships of opportunity" got too much weight. Now ogcapi loads the acronym list from the Elasticsearch synonyms set when it starts and reloads it every day. It only searches the synonyms fields when the search text has an acronym in it, so "soop temperature" is still searched as "ships of opportunity temperature" but plain words count once.

A following up PR is required to update application-staging and application-production setup.

Explain endpoint demo:

Before change, the relevance score got doubled:

[utas-raymondng]

Maybe for fuzzy tile we can do

GET /your-index/_search
{
  "query": {
    "multi_match": {
      "query": "javascrip frameworks",
      "fields": [
        "title^3",
        "title.synonyms"
      ],
      "type": "best_fields",
      "fuzziness": "AUTO",
      "prefix_length": 2,
      "tie_breaker": 0.1
    }
  }
}

and exactly match use

GET /your-index/_search
{
  "query": {
    "multi_match": {
      "query": "AI applications",
      "fields": [
        "title^3",
        "title.synonyms"
      ],
      "type": "best_fields",
      "tie_breaker": 0.1
    }
  }
}

and you just need to change inside CQLFields.java

Finally remove the existing acronym_title and acronym_desc

[utas-raymondng]

Added comments

[shaunahu]

Close this one as approach changed to: #357 (comment)