Fixes CC-5809

This covers two cases:

1. container app deployment is rejected by the coordinator api. an image has been pushed, but no app created. therefore the next time we deploy, we find the image exists already, think we can just use the previous app's image uri, but there is no prev app. this PR changes it to use the remote digest that we get from inspecting the remote manifest.
2. worker is deleted, which deletes the DO and container but not the image. if we try and redeploy with the same name, the image still exists so we hit the same problem as above. this should really be fixed on the api side (atomic deletions), but at least this unbreaks that for now.

• Tests
  • Tests included
  • Tests not necessary because:
• Public documentation
  • Cloudflare docs PR(s):
  • Documentation not necessary because: bugfix
• Wrangler V3 Backport
  • Wrangler PR:
  • Not necessary because: new