Handle Windows panics in acceptance test exit status checks

[teor2345]

Is your feature request related to a problem? Please describe.

In the zebrad acceptance tests, Windows tasks that panic seem to return the same exit code as processes that are killed by the test runner.

Edit: We discovered this issue happened because we were calling kill on a task that had already panicked.

Describe the solution you'd like

Stop calling kill on that task.

Describe alternatives you've considered

We could do nothing, and our tests would be unreliable on Windows (or our test code would be hard to understand)

Additional context

Discovered when fixing #1535

[oxarbitrage]

I think i found the cause of this bug and fixed it at df0e38d

Our acceptance test function check_config_conflict() was killing both nodes but node2 is already killed by the port conflict panic. Linux seem to dont care about this, it does not fail and the status code remains as not killed. Windows in the other hand mark the program as killed. So, i removed the killing of the second node and also the windows conditional.

[teor2345]

Great, thanks for fixing this bug, it would have confused us if we copied that code around.

[teor2345]

I've edited the task description and removed the blocker.