add jitter to the time when retries are scheduled. this is helpful in…

… our particular use case

lib/sidekiq-rate-limiter/schedule_in_future_strategy.rb

@@ -4,8 +4,16 @@ def call(work, klass, args, options)
       Sidekiq.redis do |conn|
         lim = Limit.new(conn, options)
         if lim.exceeded?(klass)
+
+          # add a random amount of jitter that is proportional to the length of time the retry is in the future.
+          # this helps us spread out the jobs more evenly, as clumps of jobs on the queue can interfere with normal
+          # throughput of non-rate limited jobs. This jitter is additive. It's also useful in cases where we would like
+          # to dump thousands of jobs onto the queue and eventually have them delivered.
+          retry_in = lim.retry_in?(klass)
+          retry_in = retry_in + rand(retry_in/5) if retry_in > 10
+
           # Schedule the job to be executed in the future, when we think the rate limit might be back to normal.
-          Sidekiq::Client.enqueue_to_in(work.queue_name, lim.retry_in?(klass), Object.const_get(klass), *args)
+          Sidekiq::Client.enqueue_to_in(work.queue_name, retry_in, Object.const_get(klass), *args)
           nil
         else
           lim.add(klass)