Fix: Address Generation Issues

[sephynox]

Summary

The SIGN_BLOCK review screens displayed addresses with the key-type prefix encoded twice and a checksum computed over the doubled bytes, so the address a user saw on device never matched their real account. The encoder now takes the raw key and key type separately, matching node-rs's format_public_key_string, and review fields split on-chain values the same way the node does.

Related Issues

• Test: Add Better Test Harness #20

Changes Made

• Review screens show the canonical keeta_ address for every field
• All three key types (secp256k1, ed25519, secp256r1) render correctly
• Device addresses now tested against node-rs and keetanet-client
• Both reference verifiers gained an address mode for those tests
• Regenerated the two sign-block review snapshot goldens

Breaking Changes

None

---

Note

Medium Risk
Changes address strings users see before signing (high user-impact) but fixes a correctness bug; scope is localized to encoding/formatting with strong reference tests.

Overview
Fixes incorrect keeta_ addresses on device by aligning encoding and review formatting with node-rs (format_public_key_string / to_public_key_string).

generate_address now takes a raw key and explicit key type byte (not Algorithm), with shared ADDRESS_PREFIX / CHECKSUM_LEN constants. GET_PUBLIC_KEY and signer derivation pass the algorithm discriminant as key type.

SIGN_BLOCK review no longer treats on-chain account bytes as a pubkey with an inferred prefix: values are split as keyType || rawKey, and 33-byte token keys default to secp256k1 (0x00) when no type byte is present. Algorithm::from_pubkey_prefix is removed.

Tests: Rust and TypeScript reference verifiers add an address mode; test_address_display.py asserts GET_PUBLIC_KEY and review-screen addresses match both references for all three key types; signature round-trip tests use shared verifier helpers in conftest.py.

^{Reviewed by Cursor Bugbot for commit 0bd1652. Bugbot is set up for automated code reviews on this repo. Configure here.}