Feature: Allow node downtime during the simulation

This issue is similar to #32.

What do we do when an individual node that we have execution permissions on goes down?
Right now, the simplest approach is just to kill the simulation.

However, there are likely some simulation scenarios where it's useful to be able to take nodes offline for a while and bring them back online. Likewise, it can be useful to continue to run the simulation with a _subset_ of the nodes that _are_ working (for example, on a signet where people shut their laptops and kill their nodes).

Rather than dying, we can add a `--keep-alive` option to the simulation which would allow us to stay up for as long as _any_ activity is successfully executing and die when we no longer have any actions that are working.