documents_v2 get populated by Aurora, mainly the /ocr/v2 endpoint and the multipart that links to this one

Theres a Job in the jobs table that tries to re-index the cache in the search collection on xbp.app/search for the collection documents_v2 this builds our search apu