Store the json intermediate db in something other than memory

[JoelHough]

So we can look at it later, or distribute it, or have it be bigger than our scraper server's memory. JSON document store or Postgres JSONB would be dandy.