Add raw_document query support to backend

[fridgepoet]

What this PR does / why we need it:
This recreates the raw_document query processing that exists in the frontend in the backend. It redirects any queries with only raw_document to the backend. It's very similar to the recent raw_data work.

The query building is based on its equivalent in Elasticsearch and the original PR is here.
The result processing is based on its equivalent in Elasticsearch. The original PR is here.

Which issue(s) this PR fixes:

Fixes #195

Special notes for your reviewer:

• As mentioned in the Elasticsearch PR, the order of the keys in the JSON will not be the same between the previous frontend implementation and this backend PR.

  results from backend are alphabetically ordered - this is done automatically while JSON.Marshall. From docs: ...The map keys are sorted. This shouldn't be an issue at all, as the order of fields in json is not used anywhere.

• Another difference is that the timestamp will be an array, not just the value, same as in the Elasticsearch PR.

  In json field (only field), @timestamp is array with value when processed trough backend (compared to just value when processed trough frontend).

  Although we can match the same behavior as the frontend by supplying only the single value from the array, if this more direct result has been acceptable for Elasticsearch users, the code will be simpler.

[iwysiu]

Mostly just style nits again, with one question about the hit fields