Previously to run a flow client code had to put together the flow,
an engine, logbook, flowdetail, and storage backend. This commit
adds two helper functions, run() and load(), so that simplest usecase
now looks like
taskflow.engines.run(flow)
Client code may also provide configuration for storage and engine if
needed, but if not needed it just works with defaults.
Engines are loaded via stevedore, as drivers in 'taskflow.engines'
backend. Now three entry points are defined in that namespace:
- 'default', for SingleThreadedActionEngine, used by default;
- 'serial', as another synonym for SingleThreadedActionEngine;
- 'parallel', for MultiThreadedActionEngine.
Closes-bug: #1224726
Change-Id: I7f4cb5c8ff7f5f12831ddd0952c202d2fd8cd6ef
cde0dee14e