Move crawl and QA logs to new mongo collection

[tw4l]

Fixes #2765

(This is a necessary part of ensuring mongo documents don't exceed 16MB. Hopefully it's also sufficient but we'll need to see in practice if there are other fields that need to be separated out.)

This PR moves crawl and QA run logs into a separate crawl_logs mongo collection.

It adds a new backend module (without a distinct API router, but the crawls module is getting quite large and it seemed to make sense to add a separate module for the new mongo collection), as well as a migration to move crawl logs from Crawl objects into the new collection. The existing nightly test for crawl error logs is fleshed out, and a new nightly test added for behavior logs.

The migration has been tested locally. I've also verified that the new collection's indices are used by the existing crawl error and behavior log endpoints.

[ikreymer]

I think it might be possible to do this entirely in mongo, might be faster using $function and calling JSON.parse?

[tw4l]

I think that it be more complicated/less efficient than it seems at first glance to handle it that way. As far as I can tell we wouldn't be able to write to both mongo collections in a single query, so we'd need to put the crawl document's log lines into memory between queries anyway. At that point, seems like we may as well keep to having a single codepath for parsing/writing log lines so we know everything's consistent.