Ae sample quality

[maffettone]

Add basic functionality for adaptive experiments based on quality with queue server. There are a handful of TODOs lingering.

1. Add sensible default args for mv and count plans
2. Validate the assumptions on documents in the index_reccomender_factory callback.
3. Update default zmq server arguments
4. Add any start or clearing of the queue necessary in the plan script.

[tacaswell]

the pdf_count plan will do both the move and the count (and get the meta-data right).

[jklynch]

Don't add None to the queueserver. If this is a case where you don't know which sample to measure and want to wait for a future recommendation just pass and tell the queueserver nothing. But maybe you do have a fallback you can do here when no recommendation is made.