Add prometheus metrics for concurrency worker

The following discussion from !139851 (merged) should be addressed:

• @DylanGriffith started a discussion:

  We rely on metrics in logs and prometheus to understand throughput of Sidekiq jobs. It could be difficult with this new defer logic to understand throughput if we are very rapidly processing a tonne of jobs and sending them to the zset.

  So I'm thinking we should add to some prometheus counter when a job is deferred so that we can subtract that from other metrics to understand the true throughput of a sidekiq worker.

  Similarly we should add a boolean of some kind to the log output to indicate a job was deferred.