Endpoint should use entrypoints for taskflow

Instead (or in addition to) of scanning classes/modules for tasks when a worker program starts up we should also make it possible to use/provide entrypoints (or a list of entrypoints) and then interact with stevedore to find and load the tasks provided by these entrypoints.

Logbook objects, not collections for taskflow

It is proposed to stop treating logbook object as collection of smaller objects: LogBook and FlowDetails should be made simple data-transfer objects, not containers for FlowDetails and AtomDetails respectively. Corresponding API should be moved to backends.

Dynamically obtain addresses and statuses from workers for worker-based engine for taskflow

Worker-based engine should be able to dynamically obtain information about workers, inluding: - service discovery: where the workers are available and what each of them can do - status: what is each worker doing (which tasks are running on it right now) This should be done in a way that does not affect, or better i...

Instead of exposing raw green threads, hide them. for OpenStack Compute (nova)

Currently there are places in nova which expose and use the greenthread code directly for various reasons. It would be nice to move to a more abstracted model where the greenthread implementation is only exposed in 'utils.py' and everything else uses modules/code there instead of using greenthread directly. This see...