Stop storing FileError items from Kingfisher Collect in the database

[jpmckinney]

open-contracting/kingfisher-collect#917 (comment)

SInce this version of Process stores the Scrapyd job ID, it's easy to use scrapy-log-analyzer to parse the log file itself. This avoids errors being introduced by Process, the network, etc.

[jpmckinney]

Hmm, the job ID is stored by the data registry. It is sent to create_collection from Collect's spider_opened callback, but the ID is not yet stored. See #341 Done

[jpmckinney]

Obviously, as part of this, we would also stop sending messages for file errors from Collect.

Edit: There is also some logic in collectionstatus that we can remove .exclude(data__has_key="http_error")

As part of this, we can delete collection_note rows WHERE note LIKE 'Couldn''t download %'

[jpmckinney]

@yolile FYI, now that we are using scrapy-log-analyzer in production, we can progress this issue, which would simplify the Collect-Process integration. (Right how, HTTP errors are showing up in Process like at https://data.open-contracting.org/admin/data_registry/job/2416/change/, but Collect should simply log these itself, and scrapy-log-analyzer can report them in the warnings email.)