Add package and module structure

[dzhoshkun]

Closes #6

• Add basic information to front page, license file and a quick start demo stub #3 should be merged in before this PR
• change base to master

[ericspod]

In general I think the architecture is structured fine with a good division between the manager and worker components. I have said in a few places that code can be made to generically handle either threads or processes using shim types and the fact that the multiprocessing primitives (lock, pip, queue, shared memory, etc.) work just fine with threads so using them everywhere doesn't hurt. I had written a bit about the buffer definition that you have and would suggest rethinking this after reviewing how Queue works and also looking into using the Pipe primitive directly. Shoving huge amounts of data down queues or pipes isn't a great idea so consider using shared memory for that.