Refactor ElasticsearchClient, splitting into separate concerns for easier testing

[npepinpe]

Description

The ElasticsearchClient is the one class in the exporter which communicates with the actual Elastic instance. Currently, it's responsibilities are:

• Configures and manage the Elastic REST client
• Network communication: (de)serailization, response/error handling, etc.
• Read templates from the resources, parse them, and substitute certain configuration options
• Compute the index name, document ID, etc., properties of each document.
• Serialization of records and templates
• Buffer serialized indexing commands in memory
• Configurable logic for when flushing should occur

I would propose to do the following:

• Extract index name and record routing logic from ElasticsearchClient #9327
• Extract index template management from ElasticsearchClient #9328
• Extract RestClient creation logic out of the ElasticsearchClient #9329
• Clean up ElasticsearchClient after previous parallel refactorings #9340
• Refactor client's unit to focus on the client's reduced responsibilities #9330
• Refactor client's integration tests #9331

The client will now only deal with network communication and in memory buffering of the documents. I would recommend moving the flushing/buffering out into its own component as well later, but the above is what provides the biggest win, I think, in terms of maintainability and test-ability.

The last two steps will be blocked by the first three, I think, but the first three can be done in parallel. Finally, refactoring the tests may include adding new testing utilities, but these will later be reused in #9321.