Tasks in graph_flow could be ignored

[Timofey Durakov]

If task in graph flow have 2 or more predecessors even if one of them is marked as ignored, task would be ignored too.
Here is example code snippet:
def allow(history):
    print(history)
    return False

# Declare our work to be done...
r = gf.Flow("root")
r_a = DummyTask('r-a')
r_b = DummyTask('r-b')
r_c = DummyTask('r-c')
r_d = DummyTask('r-d', parents_required=task.Required.ANY)
r.add(r_a, r_b, r_c, r_d)
r.link(r_a, r_b)
r.link(r_a, r_c, decider=allow)
r.link(r_b, r_d)
r.link(r_c, r_d)

# Setup and run the engine layer.
e = engines.load(r)
e.compile()
e.prepare()
e.run()

`r_d` task wouldn't be executed

To fix that is possible to add new attribute to BaseTask class, which should describe dependency between predecessors and current task execution. Possible values could be `ANY`, and `ALL`.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/208468

[OpenStack Infra (hudson-openstack)]

Change abandoned by Ben Nemec (<email address hidden>) on branch: master
Review: https://review.openstack.org/208468
Reason: This has been in merge conflict for 2.5 years. Please feel free to restore if you still want to do it though.