Harden TritRPC v1 parity: canonical IDs, decoders, repack checks, and CI/verify tooling

[mdheller]

Motivation

• Fix cross-language protocol mismatches (Rust used zeroed schema/context) so fixtures remain the single source of truth.
• Ensure deterministic AEAD AAD/tag computation and enforce full-frame parity (not just tag/payload checks).
• Add minimal structured decoders and Avro round-trip checks so fixture semantics are verifiable per-language.
• Improve repo ergonomics with a standard local make verify flow and CI that runs the same checks.

Description

• Align Rust envelope IDs with canonical constants and use them when building envelopes, and add a Rust envelope decoder that exposes AAD/tag boundaries.
• Harden Go: add DecodeEnvelope + AADBeforeTag, a TLEB3 decode helper, Avro decoders/encoders for HGRequest/HGResponse, and repack/round-trip checks in fixtures_test.go.
• Add round-trip Avro decode/encode checks and full-frame decode -> build equality assertions in Rust fixture tests, plus constant-time AEAD tag comparisons (subtle usage).
• Documentation and tooling updates: parity audit, integration readiness checklist, Makefile with make verify, and a CI workflow that runs format + tests + fixture verification; spec/docs updated to reflect port behavior (XChaCha20-Poly1305 only, AUX opaque, explicit per-frame nonces).

Testing

• Ran formatting checks (cargo fmt, gofmt) successfully as part of verification steps.
• Attempted go test ./... but it failed due to network/proxy restrictions when fetching modules (dependency download blocked).
• Attempted cargo test but it failed due to crates.io index download being blocked by network/proxy restrictions.
• Ran the Python fixture verifier but it failed locally because the cryptography package (XChaCha20-Poly1305) was not installed in the environment.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1da6dde25a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Handle TritPack243 tail markers in TLEB3DecodeLen

TLEB3EncodeLen uses TritPack243, which emits a 2‑byte tail marker (243..246 + value) whenever the trit count isn’t a multiple of 5 (e.g., any length with 1–4 base‑9 digits, including 0..6560). TLEB3DecodeLen reads one byte at a time and calls TritUnpack243([]byte{b}) while discarding the error, so if the next byte is a tail marker the decode silently drops trits and then interprets the tail value byte as a full 5‑trit group. That produces the wrong length and causes DecodeEnvelope/splitFields to mis-parse most frames.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Fix newOff advance for multi-byte lengths

off already points past the bytes read for this length, but newOff adds usedBytes on top of that (offset + usedBytes - 1 + (off - offset)). For any length that consumes more than one byte, this advances past the actual field boundary (e.g., a two‑byte length returns offset+3 instead of offset+2), so subsequent fields are misaligned and AAD/tag slicing becomes incorrect.

Useful? React with 👍 / 👎.