Get an understanding of worker capacity and throughput

[markstory]

We need to get an understanding of the throughput that an individual sentry taskworker can process. We'll need a 'busy loop' task task that runs for representational duration. We can get task duration distributions from existing celery workloads.

• How do additional child processes increase throughput and impact memory usage?
• How much CPU usage can we achieve with 4/8/16/32 child processes?