response to persist work actor failure

[lsat12357]

I've been finding works that, according to the aasm_status, failed during the persist_work stage, but are actually in Hyrax. Presumably something goes wrong after the object is saved but before the stack completes.
I think the causes I've seen so far (that aren't just problems we need to fix) have to do with the attach files job. I think we could have a job/service that retries running the attach files and checks visibility and if that operation is successful, a callback that updates the migrator work status.
We could add more services to cover other failures as they become apparent.

[lsat12357]

Add other services as needed.

[lsat12357]

because of intermittent system errors, this time around :

• ERROR: Undefined namespace prefix: /rdf:RDF/rdf:Description/dc:title/text()
• Failed to open TCP connection to fcrepo.od2-test.svc.cluster.local:8080 (getaddrinfo: Temporary failure in name resolution)
• end of file reached

a number of assets were ingested in an incomplete state.
Fixing them required some combination of:

• attaching the fileset
• setting visibility
• setting collections
• creating new sipity entity

May want to draw the line somewhere and just delete/reingest? we just had some changes to infra, likely will not have this level of error when we start migrating for real.