Add a polling_flow pattern for retrying a set of tasks up to a timeout

A common workflow pattern consists of something like this:

CreateServer()
while not server.done:
    CheckIfServerIsDone()

There should be a construct in taskflow to make such patterns easier.