sentry: add exception handling for main tasks

[canepat]

This PR introduces some additional exception handling in the following modules of sentry:

• run loop of all the Sentry main tasks (i.e. StatusManager, Discovery...): these are long-running tasks which usually should never terminate (unless the user requests a graceful exit), so here we just catch any system exception to log a critical error and relaunch it. As a result, we will see what is the root cause of the error forcing the program to exit, currently it's just a somewhat obscure co_await: operation cancelled error
• rlpx::Peer: during some tests I've found that sometimes a std::nested_exception gets raised, which we can handle properly. Moreover, I've seen Peer::post_message rarely failing to spawn Peer::send_message, thus causing the program to terminate, so I've added a warning to at least detect this situation (it seems to me like an unavoidable race between peer close and message spawing in different sub-tasks) and avoid program termination
• rlpx:: Server: again, sometimes async_accept raises a boost::system::system_error with error code boost::system::errc::invalid_argument, which we need to understand but seems to self-heal in some way