Add limit to host startup retries

There are certain scenarios where the host will attempt to startup forever, which can lead to an effectively dead host. This can happen when something in the underlying platform has gone wrong (we see this rarely with SocketExceptions on startup, for example). Instead of trying forever, giving some limit and then shutting down the process would be beneficial to letting the platform attempt to recover.

We'll need to look through all the hosting scenarios to see what the best behavior is here, but retrying forever in these cases doesn't seem like the best behavior.

Alternatively -- the host should report that it is unhealthy after some number of attempted startups so that the platform can take the correct action.

Start by mapping the hosting scenarios and startup retry paths involved in the host, then investigate how SocketExceptions during startup are handled. Define the desired behavior for bounded retries or an unhealthy status, including when the process should shut down or let the platform recover.

Written by the indexing model from the issue text.