Gorrys invariant nits

[gorryfair]

Thanks Gorry, these are all quite helpful.

1. I understand the acronyms DCID and SCID, but these aren't really
   expanded on first use in sect 5.1.

2. There's mention of "the high bit", but there doesn't seem to be any
   statement of bit order saying. Is it worth declaring network byte/bit
   order earlier in section 4? This also clarifies the placement of fields within a byte, which was basically assumed knowledge otherwise.

3. Personally, I don't think: "The length of the Destination Connection
   ID is not specified" is a great choice of words. Surely, this needs to
   be specified somewhere for a protocol instance to use it?
   This reads better as something like: The length of the Destination Connection ID is not encoded in packets with a short header and is not constrained by this specification.

4. The brackets here seem to point to connection IDs in general, not the DCID "(see Section 5.3)". Is it worth writing that as "Connection IDs are described in Section 5", rather than use brackets? The same is true for versions.

5. In section 5.3: "the wrong endpoint" is taken as the wrong transport
   endpoint, but some of our IP and subIP colleagues would use the term
   endpoint for other uses, so one insertion of "transport" would perhaps help?

6. In the appendix A, this: "QUIC forbids acknowledgments of packets" -
   still has me wondering what is actually intended to be understood,
   because the list of bullets is not guarenteed ... :

/The last packet before a long period of quiescence might be assumed
to contain an acknowledgment (it should be assumed that QUIC could allow acknowledgments of packets that only contain ACK frames)/

I'm not sure what to say here I think the combination of not... forbids threw me off the meaning.

Changes proposed in #3730