Bumping hyrax to address silent infinite retries

[jeremyf]

From the Bulkrax pull request:

Prior to this commit, when we encountered a
Bulkrax::CollectionsCreatedError, we would reschedule the same job
with a 1 minute delay. However, if the underlying issue was never
resolved, we would continually reschedule without fanfare this failing
job.

With this commit, we provide 3 logged reschedules, then if that fails
we raise an exception break the infinite cycle. By breaking the
cycle, then the upstream Sidekiq handler can determine if it should be
thrown into the Retry queue (where the jobs that raise exceptions
go).

We discovered this in Adventist watch 100 or scheduled jobs run very
fast, and apparently continually rescheduling themselves.

Related to:

• Removing ∞ silent retries of Bulkrax::ImportWorkJob samvera/bulkrax#692
• Bulkrax::ImportWorkJob silently and infinitely reschedules Bulkrax::CollectionsCreatedError exceptions samvera/bulkrax#691