Fix retry loop in NameError job recovery initializer

GoodJob's `.discarded` scope is defined as `finished.where.not(error: nil)`,
which matches any finished job with an error — including jobs with
`error_event: :retried` that already have a pending retry scheduled.

On every restart, the initializer was finding these already-retried jobs
and calling `retry_job` again. Thankfully good_job has us covered and
prevent a new retry job from being queued. This is visible with
`ActionForStateMismatchError` or `PG::UniqueViolation` errors on each
boot. Not harmful, but far from ideal.

Symptoms visible in the logs on every restart:

  Failed to enqueue job for retry SomeJob (job id: b03926e2-...):
  PG::UniqueViolation: ERROR: duplicate key value violates unique
  constraint "good_jobs_pkey"
  DETAIL: Key (id)=(a798492e-...) already exists.

Exclude jobs with `error_event: :retried` so only truly stuck jobs
(unhandled, retry_stopped, discarded) are re-queued.

config/initializers/retry_failed_jobs_with_name_error.rb

@@ -42,6 +42,7 @@ Rails.application.configure do
 
     GoodJob::Job
       .discarded
+      .where.not(error_event: GoodJob::ErrorEvents::RETRIED) # reject discarded jobs that were already retried
       .where("error LIKE ?", "NameError: uninitialized constant %")
       .filter do |job|
         # Only retry jobs with NameError related to the job class name.