Make taskworker worker support concurrent work

[markstory]

The current worker implementation is single threaded and will have limited throughput. Expand the worker to support concurrent processing based on CLI flags. --concurrency=12 should create 12 worker threads/processes allowing a worker to process up to 12 tasks concurrently.

Rough thinking would be to have a single thread responsible for fetching tasks and sending rpc responses back and a pool of worker threads that execute tasks. Processing timeouts could be tricky to enforce though as python threads don't have a way to do that. Perhaps celery worker has some tricks we could replicate.