Better datasource abstraction

[richard-jones]

• PR: [when the PR is created, put a link to it here. DO NOT use the "linked PR" feature]

---

We want to be able to support multiple data sources, including:

• Multiple ES versions
• OpenSearch
• SOLR
• Static CSVs

Right now the code is capable of this to some degree, but a more coherent abstraction between the internal state and the queries to the datasources and back again with their result lists would be sensible.

Also, to be able to support multiple datasources in parallel.

Acceptance Criteria

• It's easy to integrate new datasources
• Be able to integrate two or more datasources at the same time