[noise] Potential cleanups for a rev35?

[Andris]

> I don't think we need a protocol change,
> "transport" messages play a special role in completing the handshake.

That certainly would be helpful. I would, however, caution against calling transport messages "tokenless messages", since it appears that many intuitively understand "tokenless messages" in the noise context to mean handshake messages without tokens. Since such messages can perform a useful role in some handshake patterns and the way they are handled does differ from transport messages in some crucial ways, I believe a less easily misunderstood terminology would prove more useful.