For now "success" means "not failure", which introduces the least surprise for tests that have
an assumption violation / failure. Tests that fall into this category are still considered
"successful" by buck, though other parts of the system (specifically, event listeners) can do
differently if they please.