HMS-5202: don't always rerun failed snaps

[jlsherrill]

Summary

After some investigation it was discovered that currently all failed snapshots are being re-run every hour. This results in no 'extra' repos being synced. This change results in failed snapshots only being re-run every day.

This also fixes a couple of things, such as sorting by finished_at (instead of queued_at), and re-snapshotting every 25 hours instead of 24 (need to subtract one from the query)

Testing steps

1. Create 3 repos with snapshotting, two that are real and one that isn't:

https://fixtures.pulpproject.org/rpm-unsigned/
https://jlsherrill.fedorapeople.org/fake-repos/needed-errata/
https://redhat.com/notarepo

2. on main, run go run cmd/external-repos/main.go nightly-jobs 24

for consistentcy do not re-run it until snapshots are done
everytime you run it, you'll see one snapshot get run and its always the same one, the one that failed

3. re run go run cmd/external-repos/main.go nightly-jobs 24 on this PR,
   now everytime you run it, you'll get a different one of the repos which should alternate between all 3:

i added some logging to make this easier to see: enqueued snapshot for repository config 147df40e-e1be-40a5-bcb2-726b6e63818c

[jlsherrill]

[test]

[jlsherrill]

https://issues.redhat.com/browse/HMS-5202

[rverdile]

looks good!

[jlsherrill]

This actually doesn't require qe