Consider extracting exporter to Python package

[jpmckinney]

e.g. Kingfisher Export ("Exports OCDS data as JSON dumps from Kingfisher Process").

It can implement the two workers (export, wipe) and otherwise offer functions to interact with the filesystem. At a later date, we can consider pre-computed Spoonbill exports (at least CSV for large datasets) and other export formats (like SQLite, pg_dump, etc.).

[jpmckinney]

Re: exporting data from Kingfisher Process, it's only needed by the data registry, so I think it's fine to leave the logic here. We can extract it later if we find a use case.

viz #314 #292 and #304, we're not sure if we'll forever have Kingfisher Process as the source of exports. So, it is also premature to extract this logic.