Store failed and non-build jobs in the analytics database

[jjnesbitt]

This PR updates our process_job function and all child data gathering functions to operate on failed jobs, as well as other job types like generate, etc. This implementation has the following effects:

1. The "type" of a job is now stored in the database (build, generate, etc.)
2. I've created some util functions for retrieving gitlab data, due to our current cluster timeout issues.
3. The gitlab section timers are now stored in the JobDataDimension
4. I've created a script to migrate existing LegacyJobAttempt rows to the current format. This must be run manually, after this is deployed, as it's lengthy and not suitable to be run automatically in a Django migration.

I've tested both the regular process_job workflow on various job types with real webhook payloads taken from Gitlab, as well as the job migration management command. Both seem to work well.

This should remove the last vestiges of the Legacy models that are used as a part of the job processor. After this is merged, we can check that none of the legacy data is unaccounted for or otherwise still being relied upon, and make a follow on PR to finally remove them.