Improve tailnet logging

Improve `tailnet` logging, allowing Coder support and Coder admins to easily distinguish normal versus abnormal networking events

* Include descriptions of Peers that we tunnel to. It’s often the case that there are multiple connections being made in an agent log file, and we’re not sure which one the customer or user was complaining about. This will require protocol changes to send this information.
* Improve logging so that we always include the node-key identifier when logging about Peers. There are two different identifiers used (for historical reasons), and so it’s hard to search the file and get the whole story.
* Ensure that INFO logs tell the important connection events including establishment of the tunnels and disconnections.
* Improve messages to make it clear that “disconnection” is a graceful disconnect, and “lost” is an unexpected thing.
* Log at WARN if peers are lost.