Support streaming the results of generators

[WardLT]

Add the ability for each element from a generator to get sent back to the thinker before the task finishes.

Streaming is enabled by adding a queue to the PythonGeneratorMethod wrapper. The wrapper will use the queue to send intermediate results back to the Thinker while executing the generator.

• Check in with team if they like the new teminology
• Add example application showing how to use the features
• Documentation for streaming, which includes warnings that PipeQueues doesn't work on distributed systems
• Make submitting the task queue asynchronous [Wait until later]