Shut down the entire node when their is a consensus panic

[cmwaters]

Currently a panic seems to only stop the consensus reactor but keeps the rest of the reactors running. This can give node operators the false impression that their node is still working. It might be more helpful to fail more loudly by stopping the entire node.

This was brought up when trying to run an E2E test. The node panicked but from the RPC client's perspective, the error that they were getting was just:

tx was evicted from the mempool

(the tx being evicted because the consensus reactor wasn't running anymore