Editorial nits in quic-tls draft

[martinduke]

Some things that were unclear when I tried to cleartext encryption from the -07 draft. I am more or less a TLS novice, so YMMV:

1. Section 5.2.1: For the cleartext secret, it would be useful to state quite clearly here that the connection ID is the "wire image" or "network order" version of connection ID; implementers should not change to host order, etc before calling HKDF-Extract.

2. Section 5.2.4: "the IV length is the larger of 8 or N_MIN (see section 4 of RFC 5116)". This is super user-unfriendly. At a minimum, we should state that the N_MIN for AES_128_GCM (needed for cleartext) is 12 bytes. At best, we should just list the IV size for supported ciphers.

3. Section 5.3: "Prior to TLS providing keys, no record protection is performed and the plaintext, P, is transmitted unmodified." I believe this paragraph is obsolete.

[martinthomson]

The wire order thing is being hashed out in #833. An opaque value can only ever be serialized one way, but if we decide that it's a number, then we will merge #881.