quic wire protocol

[majestrate]

we have a home grown wire protocol that we should replace with quic tls1.3.

[majestrate]

i dont think the internals refactor is blocking this in practice. it's far away enough from what i am working on now that it can be done without hassle.

[majestrate]

updating the wire protocol to quic will let us get rid of the jenky old one which:

• has a dumb cpu bottleneck in the hot event loop
• has a state machine race condition in the handshake
• has dubious code quality
• wasn't well designed

ietf quic with tls1.3 has:

• multi threaded cryptography
• well defined state machine with 0 rtt resume
• written by people just doing this one thing, offloads work of making it nice to them.
• drafted by ietf and various professionals outside of google

[majestrate]

we started work on a libquicinet for both wire protocol and embedded lokinet