Handle duplicate node names & support UUID based proxying (Version 2)

[fspmarshall]

This PR implements an alternative strategy to the original UUID based routing strategy seen in #3340. Rather than explicitly routing by UUID, this implementation checks if the supplied hostname appears to be a UUID and guesses the intended behavior:

$ tsh ssh [email protected]
error: ambiguous host could match multiple nodes
  
Node  Name  Node ID                              Address        Labels
----------- ------------------------------------ -------------- ---------
example.com ae79f375-633e-4a81-941e-83faf4a3c966 127.0.0.1:3022 foo=bar
example.com 17a980a9-ce32-4c95-a5c4-c4b4717851e9 127.0.0.1:4022 spam=eggs            
  
Hint: try addressing the node by unique id (ex: tsh ssh user@node-id)
Hint: use 'tsh ls -v' to list all nodes with their unique ids

$ tsh ssh alice@17a980a9-ce32-4c95-a5c4-c4b4717851e9

By treating UUIDS as hostnames, this PR is (mostly) much simpler than #3340. There is, however, one major exception: because clients treat UUIDs as hostnames, certificates now need to include UUIDs. This PR is currently a WIP since it does not yet include any mechanism for forcing nodes to get their certificates re-issued with UUIDs.

Fixes #2396

[webvictim]

Assuming we go with this approach, we should add something to the docs (probably here: https://gravitational.com/teleport/docs/cli-docs/#tctl-auth-rotate) describing what an in-place rotation is, what it does, what it doesn't do and when you might want to use it.

[fspmarshall]

Assuming we go with this approach, we should add something to the docs (probably here: https://gravitational.com/teleport/docs/cli-docs/#tctl-auth-rotate) describing what an in-place rotation is, what it does, what it doesn't do and when you might want to use it.

The rotate-in-place strategy has been shelved for now. We're opting to have nodes check their certs on startup instead.

[benarent]

👍 Hints look good.

[fspmarshall]

retest this please

[benarent]

I left a few comments. I know in the past we've used server, but we should aim to use node/nodes vs servers for consistently. It's most important for log messages and user facing input.

[fspmarshall]

retest this please