Formalize data availability, liveness, safety, synchrony property definitions

[cwgoes]

(not yet proofs - that's #42)

In various locations throughout the spec, "data availability", "liveness", "synchrony", and "safety" are used to mean subtly different things, which (although the referents in question are all salient) is confusing and may lead to misinterpretation. These three terms should be clearly defined and further disambiguated where necessary (e.g. liveness of timeouts vs. liveness of packet confirmation, safety of timeouts vs. safety of exactly-once, data availability to relayer vs public).

Then use these definitions elsewhere, replacing "guarantee" or the like where present.

[cwgoes]

• Get rid of "public" / "private", discuss instead what data is available to who when
• Relayer could be carrier pigeon (relayer still “untrusted’, but enabled by giving the paper to the carrier pigeon)
• Relayer constrained by data availability & anti-spam on receiving chain, liveness dependent on at least 1 correct relayer satisfying both