Send hashes in the echo round instead of the full messages

[fjarri]

Some echo broadcasts in CGGMP'24 include large structures that scale linearly with the number of nodes, so the corresponding echo round messages would scale quadratically. This PR replaces sending full payloads the second time with just sending payload hashes. The trade-offs are:

• every message needs two hash applications instead of one (hash the payload, then hash (payload hash, metadata))
• more information to keep in the Evidence (need to keep the echo messages themselves and the signed hashes)

Also fixes a bug with the TestHasher OutputSize and buffer.len() mismatch. Before this PR the created digest was actually ignored when creating TestSignature, so the bug didn't fire.

This approach hashes in two stages when it's technically not necessary (that is, for non-echo messages), but I don't think it leads to any noticeable performance decrease — with most hashes it's the update part that's slow, not the finalize. So the only additional cost is an extra tag pushed into the digest.

[coveralls]

Pull Request Test Coverage Report for Build 13193469924

Details

• 144 of 158 (91.14%) changed or added relevant lines in 7 files are covered.
• 3 unchanged lines in 3 files lost coverage.
• Overall coverage increased (+2.7%) to 71.202%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
manul/src/protocol/message.rs	17	18	94.44%
manul/src/session/transcript.rs	17	18	94.44%
manul/src/session/evidence.rs	33	39	84.62%
manul/src/session/message.rs	61	67	91.04%

Files with Coverage Reduction	New Missed Lines	%
manul/src/session/evidence.rs	1	53.0%
manul/src/session/message.rs	1	93.91%
manul/src/protocol/message.rs	1	80.43%

Totals
Change from base Build 13062498128:	2.7%
Covered Lines:	1978
Relevant Lines:	2778

---

💛 - Coveralls

[dvdplm]

LGTM

EDIT: Although I have a question. This optimization is CGGMP-specific and I wonder how well it jives with manul's goal of being a general purpose "round manager" library. I'm fine merging CGGMP specific stuff but it's worth keeping track of where we stray from the general-purpose goal.

[fjarri]

This optimization is CGGMP-specific

I think it's a pretty general thing, we don't want to send whole messages the second time. Even without quadratic scaling echos with full messages can get pretty large.

[fjarri]

In addition to the requested changes, also added the forgotten length prefix when hashing variable-length bytestrings. Tend to forget this kind of stuff because in synedrion it happens automatically.