Make RunnableSummary.is_intermittent return None when there is not enough information

[marco-c]

RunnableSummary.is_intermittent should return None when we don't have enough info to tell for sure (a bit like we do for cross-config and config-consistent failures).
The prior regressions detection logic (

mozci/mozci/push.py

Lines 964 to 988 in 944248b

	if name in runnable_summaries:
	found_in_parent = True
	summary = runnable_summaries[name]
	# If the failure is not intermittent...
	if not failure_summary.is_intermittent:
	# ...and it failed permanently in the first parent where it ran, it is a
	# prior regression.
	# Otherwise, if it passed or was intermittent, it is likely not a prior
	# regression.
	if (
	summary.status != Status.PASS
	and not summary.is_intermittent
	):
	prior_regression = True
	break
	# If the failure is intermittent and it failed intermittently in a close
	# parent too, it is likely a prior regression.
	# We need to explore the parent's parents instead if it passed or failed
	# consistently in the parent.
	elif summary.is_intermittent:
	prior_regression = True
	break

) should take None into account and handle all possible cases: parent None child None, parent intermittent child intermittent, parent real child real, parent None child intermittent, etc..

E.g. in a case like this:

def test_intermittent_failure_and_consistent_in_parent(create_pushes):
    """
    Tests the scenario where a task failed intermittently in a push, and failed
    consistently in its parent.
    """
    p = create_pushes(3)
    i = 1  # the index of the push we are mainly interested in

    p[i - 1].tasks = [
        create_task(
            id=fake_id(1),
            label="test-prova",
            result="failed",
            classification="not classified",
        ),
    ]
    p[i].tasks = [
        create_task(
            id=fake_id(2),
            label="test-prova",
            result="failed",
            classification="not classified",
        ),
        create_task(
            id=fake_id(3),
            label="test-prova",
            result="passed",
        ),
    ]

How do we know if the task wouldn't have failed intermittently in the parent push too, had it had a chance to run multiple times? We should consider this case as a prior regression (assuming we have no classifications and no backout information here, the situation is different in case we have them).