test: add golden test for hash_proto

[iamquang95]

The golden file and output are generated by charon. Then test the input/output on pluto side.

[iamquang95]

@claude review PR

[claude]

Claude finished @iamquang95's task in 2m 9s —— View job

---

PR Review: test: add golden test for hash_proto

• Read the diff (msg.rs test, golden JSON, Cargo changes)
• Cross-check against charon source
• Verify dependency wiring
• Post review

---

Overall this is a clean, well-motivated test-only change. It replaces the single inline-constant test (hash_proto_matches_seeded_unsigned_dataset) with a table-driven golden-vector test sourced from charon. Nice work. ✅

What's good

• Faithful cross-impl anchoring. The attester_seed0 vector's hash d8f9bc3d… is exactly charon's own TestHashProto golden (charon/core/consensus/qbft/msg_internal_test.go:34), so a green test proves both that pluto's hash_proto matches charon and that the captured vectors are faithful.
• Logic parity confirmed. hash_proto (msg.rs:272) mirrors charon's hashProto (msg.go:140): deterministic marshal → PutBytes → Merkleize → HashRoot, plus the same Any-rejection guard. The BTreeMap keying matches charon's deterministic map marshalling.
• Good coverage spread. Single-DV, multi-DV cluster shape, and the empty-set boundary (all-zero root) are all exercised, plus distinct UnsignedData byte shapes across attester/aggregator/sync_contribution.
• Drift guard. Pinning charon_ref == "v1.7.1" and asserting non-empty vectors is a nice guard against silently editing the golden file.
• Dependency wiring is correct. serde is added to [dev-dependencies] (not runtime deps), and the workspace serde already enables the derive feature — so #[derive(serde::Deserialize)] compiles.

Minor observations (non-blocking, nits)

1. duty is purely descriptive. hash_proto hashes the raw encoded bytes and is duty-agnostic — the duty/name fields only feed the failure message, not any behavior. The real coverage value comes from varying data_hex, not the duty label. That's fine (it documents intent), just worth being explicit that this test does not exercise any per-duty branching.
2. empty_set → zero root. Worth a one-line comment in the JSON note (or a // charon-verified: empty encoding merkleizes to the zero root in the test) so a future reader doesn't mistake the all-zero hash for a placeholder/unfilled value.
3. Inline struct defs. Declaring Entry/Vector/File inside the test fn is fine for a one-off; if more golden-file tests get added later, consider lifting them into a small shared test helper.

Note on verification

I could not run cargo test -p pluto-consensus in this environment (command approval not granted), so I validated by source inspection and by confirming the attester_seed0 vector equals charon's committed golden. Please make sure CI runs the suite before merge — the empty-set and multi-DV vectors in particular are the ones that would surface any real divergence.

_{· branch feat/test-strategy-app-run}