✨ [CW-28621] feat(ada): support putting token assets into change utxo for all tx types

[a0979470582]

Summary

Lets users spend ADA from UTXOs that contain Cardano native tokens. Previously these transactions failed (ValueNotConservedUTxO) because the change output couldn't carry the tokens back. This makes every ADA transaction type's change output token-aware through a single unified encoding — no dedicated token tx type.

• Each change output carries a pre-encoded CBOR value: a bare uint for ADA-only change, or 82 <lovelace> <multiasset> when the change holds native tokens (canonically ordered — keys sorted by length then bytewise)
• New ChangeOutput type (the change field + optional assets?: TokenAsset[]); the app passes structured token data, the SDK does the CBOR encoding
• Single getChangeArgument (value-blob, 292 bytes) shared by Transfer and all six staking types; pure-ADA tx bodies stay byte-identical, so existing signatures and flows are unaffected
• Paired card-side change: coolwallet-signing-tools#201

Design notes

• Superseded approach: first built as a dedicated TransferWithTokenChange tx type to isolate risk, then merged into the existing types — the token change layout is a strict superset of pure ADA, so unifying drops a whole type and removes drift between the token and non-token paths.
• 6AE2 on-card throw: the change-value length field originally used setBufferInt(.., 0, 200), but the card's 1-byte range comparator can't take an upper bound ≥ 128, so it threw for every value. Switched to setBufferIntUnsafe (the SDK already caps the value at 200 bytes in getChangeArgument).
• Canonical CBOR: multi-asset map keys must be sorted (policy id bytewise; asset name by byte length then bytewise) or the node rejects the tx — buildMultiAssetCbor enforces this.

Status — verified on real card ✅

• All 7 scripts regenerated for the unified encoding and re-signed with CBPubKey
• Signing matrix test: every tx type × {without token, with token}, all green on a real card with no -u
• The two preserved pure-ADA baselines (StakeRegisterAndDelegate / Abstain) pass unchanged → confirms pure-ADA tx bodies are byte-identical under the new encoding (zero regression to existing flows)
• getTransactionSize covered for the token path (app fee estimation); every with-token case is a consistent +80 bytes over its no-token pair

Remaining (out of this PR's scope)

• Testnet acceptance: submit one real value-conserving token tx on preprod/preview to confirm end-to-end on-chain (validates minADA, value/token conservation, canonical ordering together)
• App side (handled separately, see app PR #10212): minADA calculation, enumerating every token in the selected input UTXOs, value conservation

Known limit: the change value blob is capped at 200 bytes (~5 tokens); beyond that the SDK throws in JS rather than emitting a malformed tx.

🤖 Generated with Claude Code

PR Summary by Typo

Overview

This PR introduces support for including token assets in change UTXOs across all Cardano (ADA) transaction types. This enhancement allows for more complex transactions involving native tokens to be handled correctly.

Key Changes

• Updated package version to 2.0.2.
• Modified transaction scripts and signatures for various ADA transaction types (e.g., Transfer, StakeRegister, Delegate) to accommodate token assets.
• Introduced new TokenAsset and ChangeOutput interfaces, and updated RawTransaction to utilize ChangeOutput for handling token-bearing change UTXOs.
• Implemented buildMultiAssetCbor and encodeOutputValue utilities to ensure correct and canonical CBOR encoding of multi-asset values.
• Expanded transaction tests to cover scenarios with and without token assets in change outputs, and added dedicated tests for multi-asset CBOR encoding.

Work Breakdown

Category	Lines Changed
New Work	328 (76.6%)
Churn	23 (5.4%)
Rework	77 (18.0%)
Total Changes	428

To turn off PR summary, please visit Notification settings.

[typo-app]

Static Code Review 📊

✅ All quality checks passed!

[typo-app]

AI Code Review 🤖

Files Reviewed: 10
Comments Added: 4
Lines of Code Analyzed: 505
Critical Issues: 0

PR Health: Good ✅

_{Give 👍 or 👎 on each review comment to help us improve.}