Skip reaping when queues are too large

[francesmcmullin]

Is your feature request related to a problem? Please describe.
This is a follow up to an observation in #663, when sidekiq queues are very long (e.g., more than 2,000 jobs) the reaper can take a very long time to run, more than likely hitting its timeout (depending on config).

Describe the solution you'd like
For applications with traffic/activity spikes, sidekiq queues can get very full, but they are eventually emptied. It's fine if reaping is postponed until queues are lower and there's more slack in the system.

Describe alternatives you've considered
We could play with the reaper timeout and count settings to try and find a balance such that reaping can continue even under heavy load, but for our use case it's not a problem if reaping is delayed during heavy activity.

Additional context
All my observations relate to the ruby reaper, it is possible that the lua reaper has no trouble with big queues

[mhenrixon]

Thank you for the issue, I'll do some thinking on this but I will likely not be able to achieve much until after Christmas.

I am moving my family across Germany on Monday.

[francesmcmullin]

Of course! You've helped a huge amount already. I hope the move goes really well. Thank you =)