Add mapping of controlled vocab fields

[danamansana]

Adds logic to map free text queries against controlled vocabulary fields to the corresponding controlled vocab fields.

The basic logic is that a match between a term and a document means that the document contains any of the ids that contain the term, or whose corresponding labels include the term, except in case of == where we require that the id is exactly the query term

For adj/=/== queries, "term" means the whole query. For any/all queries, "term" means any word in the query, and a match means any/all terms are a match according to the above criteria

Overview of code changes:

• Adds controlledVocabularies config file (eventually we'll want to look into automatically keeping this up to date with Research Catalog, but it doesn't change much)
• buildAtomicMain now switches between regular queries and controlled vocab fields, which have to be mapped first. Also scope has to be passed down
• adds the mapping function
• Adds tests and fixtures.

Also, an unrelated change to switch the shelfMark field we are using to the text version, to accommodate some requests to be able to match a part of a callnumber query

[charmingduchess]

Instead of hard coding these values, we should rely on nypl core values where possible. Something like what happens in this file that generates these values from nypl-core mostly. I don't think we need to be running the aggregations queries since we're not displaying these values to users. But we should be generating values dynamically from nypl-core - especially or at least the collection/holding location mapping.

The language mapping isn't in nypl core, but is served in JSON-LD from loc here. That one maybe makes sense to just hardcode here, but it ought to have some way of dynamically regenerating it. Personally, I wouldn't be opposed to accessing it from the loc.gov endpoint. This iso language code library might also be worth looking into.