Geo: Increase progressive backoff cap for sync retries of things which are missing-on-primary
A Geo secondary can be in a state where very many things are marked missing-on-primary. For legacy data types, these are internally treated as failures, in that they are retried in a similar manner.
But progressive backoff is capped at 1 hour, which is fairly short-- geo.staging.gitlab.com has 91k uploads/artifacts/lfs objects being retried every hour.
Proposal
Increase progressive backoff cap for missing-on-primary retries to 4 hours.