feat: resolve a DID's current XCH payment address from chain (resolve_xch_address)

[MichaelTaylor3d]

What & why (Closes DIG-Network/dig_ecosystem#1239, adopts #1240)

Two coupled changes in one unit of work (§1.3b):

1. resolve_xch_address — the money-critical DID→XCH payment-address primitive. Walks the DID
   singleton to its current authenticated tip and returns the tip owner's payment Address
   (tip.info.p2_puzzle_hash, bech32m under a caller-supplied prefix — network-agnostic, no
   hardcoded HRP). Chain-READ only: never signs, never broadcasts (INV-1/INV-2).
   • resolve_xch_address(launcher_id, prefix, source) -> DidResult<Option<Address>>
   • resolve_xch_address_from_did_string(did, prefix, source) -> DidResult<Option<Address>>
2. Migrate the ChainSource seam onto the canonical dig-chainsource-interface v0.1.0 (crates.io):
   dropped dig-did's LOCAL ChainSource trait + SingletonLineage, depend on the published interface,
   re-export both so dig_did::ChainSource / dig_did::SingletonLineage are unchanged (additive).

New public surface: resolve_xch_address, resolve_xch_address_from_did_string, DidError::LauncherMismatch
(additive), and a re-exported chia_sdk_utils::Address.

Money-critical soundness (NC-9) — the parent-walk guard

A wrong resolution silently pays the wrong recipient, so the resolved address MUST be provably bound to
the REQUESTED identity. The cheap tip.info.launcher_id() == launcher_id check is INSUFFICIENT — that
curried launcher id is attacker-chosen (the pay_to_coin_wearing_a_singleton_puzzle_hash class). After
walk_did_lineage_to_tip, resolve runs authenticate_singleton(tip.coin.coin_id(), source) — which
walks the parent-spend chain to the GENUINE launcher — and requires that authenticated launcher to
equal the requested launcher_id, else DidError::LauncherMismatch. NC-9 (on-chain data requires
on-chain proof, fail-closed) is satisfied by this parent-walk authentication, not by trusting any
attacker-chosen field. Proven load-bearing: neutralizing the guard makes the money test fail (resolves
to the attacker's address).

NC-8 is N/A — resolve_xch_address is a chain READ that writes nothing on-chain (no spend, no
broadcast).

Fail-closed None/error table (implemented exactly)

unlaunched/melted → Ok(None); malformed did:chia → InvalidDidString; tip creating-spend absent →
NoIdentitySingleton; tip not a DID → NotDid; spoofed curry → fail-closed (NotDid/NotASingleton);
genuine launcher ≠ requested → LauncherMismatch; chain read fails → Chain; over-deep → LineageTooDeep;
Address::encode fails → Parse.

Tests (TDD, adversarial)

38 tests pass. The 6 mandated adversarial tests: resolve_rejects_an_echoed_different_dids_tip (THE
money test — dishonest source echoes an attacker DID's tip for the victim launcher → LauncherMismatch,
and asserts the result is NOT the attacker's address), resolve_rejects_a_spoofed_curry_singleton,
resolve_returns_none_for_unlaunched_or_melted, resolve_from_did_string_rejects_malformed,
resolve_happy_path_matches_the_owner_address, resolved_address_roundtrips_through_decode (+ a
did-string happy-path). Uses the interface's MockChainSource (testing feature) + a Simulator-backed
source for genuine singleton spends. No live network.

Blast radius checked

ChainSource trait callers are all in-crate: authenticate_singleton[_bounded],
walk_did_lineage_to_tip, prove_lineage, and the test sources. The migration re-exports the interface
trait/type so the dig_did::* paths are preserved (additive externally — new crate, no external
implementors yet). resolve_xch_address is net-new (no callers). Risk: LOW-MEDIUM (the trait widens
required methods, inherent to adopting the canonical interface per #1240; existing lineage tests updated
to implement the full trait).

Quality gates (local)

• cargo fmt --all -- --check clean · cargo clippy --all-targets --all-features -- -D warnings clean
• cargo llvm-cov nextest --all --fail-under-lines 80 → 91.67% lines / 89.01% regions
• RUSTDOCFLAGS=-D warnings cargo doc --no-deps clean · version bumped 0.3.0 → 0.4.0 (minor: additive)

Do NOT merge — handing back to the orchestrator for the triple gate.

[MichaelTaylor3d]

PASS — correctness/quality gate (independent fresh-context review)

Money-critical resolve_xch_address primitive reviewed against design #1239 (comment 5016826175). Verdict: PASS.

Money-guard load-bearing: CONFIRMED. Neutralized the launcher check (if false && authenticated.launcher_id != launcher_id) — resolve_rejects_an_echoed_different_dids_tip fails (resolves Ok(Some(...)) for the victim launcher instead of LauncherMismatch). The guard at src/resolve.rs:335-338 runs authenticate_singleton (parent-spend walk to the genuine launcher) AFTER walk_did_lineage_to_tip and BEFORE building/returning the address — correctly ordered, genuinely load-bearing, non-tautological.

Fail-closed: CONFIRMED. No path returns a wrong address or spurious Ok(None). Tip reconstructed from parent's genuine singleton successor (coin-id verified, resolve.rs:268-270); address validated via .encode() before return; every wrong case maps to a typed Err (LauncherMismatch/NotDid/NotASingleton/NoIdentitySingleton/LineageTooDeep/Chain/Parse). Spoofed-curry resolving to NotDid (vs NotASingleton) is fail-closed — design-table doc nit only.

Interface adoption, no regression: CONFIRMED. Local ChainSource/SingletonLineage dropped, re-pointed to dig-chainsource-interface = "0.1" (crates.io, no git dep, no [patch.crates-io]); dig_did::ChainSource/SingletonLineage paths preserved via re-export (resolve.rs:47, lib.rs:78-80). Lineage/auth walk compiles + all lineage tests green.

Gates: cargo test 38/38 green (36 lib + 2 golden); clippy -D warnings clean; fmt clean; coverage 91.67% lines / 89.01% regions (≥80%). Docs: SPEC §3 U10 + §6 LauncherMismatch, README, CHANGELOG, lib.rs re-exports (Address + resolve fns) all present. Version 0.4.0, additive. No cruft.

No blocking findings. No pre-existing review threads. Orchestrator: clear to merge on green checks + zero unresolved threads.