Opinions index changes to improve MLT queries

[albertisfu]

Following comment #4305, this PR adds a new field to the Opinion Index mapping. It adds a field called combined_fields that combines multiple fields used to look for similar documents into a single one. This field uses the english_exact analyzer to avoid removing duplicates and avoid synonyms that could impact the quality of the query. Additionally, the field uses term_vector to only store fields, improving the performance of MTL queries as recommended in the documentation.

To test this approach, we can:

1. Merge the PR.
2. Within a maintenance pod, rename opinion_index = Index("opinion_index") to a different name, for instance "opinion_index_combined", in es_indices.py. This ensures the original production index won't be affected.
3. Create the new index by running manage.py search_index --create --models search.OpinionCluster.
4. Finally, perform a reindex:

POST
/_reindex?wait_for_completion=false

{
   "source":{
      "index":"opinion_index",
      "size": 1000000 # Remove this line to re_index all the documents.
   },
   "dest":{
      "index":"opinion_index_combined"
   }
}

This will return a task_id that we can use to monitor the task progress.

If we only want to assess the quality of results, reindexing a subset of the index may be sufficient. However, if we also want to assess the performance of this approach, indexing the entire index would be better.