SOCKS handshake faux username/password value-space

We (ab)use the SOCKS username/password field for passing isolation information. In the RPC system we also propose (ab)use it for method-invocation-including-datastreams-via-socks. !2373 (comment 3071231)

Problems:

1. I don't think we document the value-space assignments and their semantics
2. I'm have qualms about the value-space assignments in !2373 (merged) (RPC)
3. When we introduce new value-space assignments eg for RPC (!2373 (merged)) we should make sure that our allocated (username,password) values cannot be misinterpreted by existing software. That probably means updating other Tor SOCKS5 proxies to have them reject values that we're changing the meaning of.
4. We need a plan that leaves space for future expansion without risk of confusion

Some of these ovrerlap. We should address this before we start to invite people to use the RPC system.