Test more edge cases and failure modes

[raffomania]

I really like the test setup and the happy path tests are a good start. Here are some more edge cases that would be really useful to test, preparing us for the inevitable hetzner failures:

• What happens when consul crashes, needs to be restarted or becomes unavailable in any way? Should we just give up, or should we add a retry logic for recovery?
• Same if the network is not reachable when starting (see recent hetzner issues). We can't do anything there and I wouldn't expect wiresmith to do anything but crash. It would still be good to cover this case in tests, if only for documentation purposes.
• Nodes that become unreachable are removed from other peers after a certain timeframe. This should be tested.