feat: core derive_wallet_id for hardware watch-only wallets

[coreyphillips]

Summary

Subtask 1 of #107. Core now owns the wallet_id derivation for hardware (watch-only) wallets so iOS and Android produce the same id for the same physical device by construction, instead of each platform reimplementing it. The Swift side currently improvises HwWalletId.derive = "trezor:" + SHA256(sorted xpubs) and flagged it as interim pending agreement with Android — if the two platforms derive independently they drift and activity won't reconcile cross-platform.

What changed

• New #[uniffi::export] pub fn derive_wallet_id(device_type: String, xpubs: Vec<String>) -> String.
• Device-agnostic (device_type keeps trezor / ledger / … distinct), so it generalises beyond Trezor.

Canonical derivation

1. sort xpubs lexicographically (input order doesn't matter),
2. join with a single \n,
3. SHA256 the UTF-8 bytes, hex-encode lowercase,
4. return "{device_type}:{hash}" (e.g. trezor:ab12…).

Trezor watch-only is new, so core defines the canonical form; platforms call this instead of rolling their own.

Tests

cargo test derive_wallet_id — order-independence, device_type distinctness, a fixed known-vector, and the empty-xpubs case.

🤖 Generated with Claude Code

[ben-kaufman]

Two things I think we should tighten before merging:

1. derive_wallet_id should reject an empty xpubs list instead of returning trezor:<sha256 empty string>. Otherwise any caller that reaches this with no account xpubs gets the same wallet id for that device type, which can collapse separate setup failures into the same activity scope. Could we make this return an error and update the empty-input test?

2. The bindings look stale or partially regenerated. The Python wrapper references uniffi_bitkitcore_fn_func_derive_wallet_id, but the committed dylib does not export it, so python3 -c "import bitkitcore" fails with symbol not found. The Swift wrapper also calls the new symbol while the iOS headers and XCFramework do not include it, and Android has no deriveWalletId wrapper or JNI declaration. Could we regenerate and commit all platform bindings and artifacts from this Rust head?

[coreyphillips]

Thanks @ben-kaufman, both addressed in 80dadad.

1. derive_wallet_id now returns Result<String, ActivityError> and rejects an empty xpubs list (also blank entries and a blank device_type), so an empty set can no longer collapse separate devices into one activity scope. The empty-input test was updated to assert the error.

2. Regenerated all platform bindings and native artifacts from this Rust head via build.sh all: Swift + bitkitcoreFFI.h + the XCFramework, Kotlin + the jniLibs .so files, and the Python wrapper + libbitkitcore.dylib. The new export is now present and callable in all three, so import bitkitcore no longer fails with a missing symbol.

[ben-kaufman]

Looks good, thanks for fixing the empty xpubs case and regenerating the bindings.

One release note before publishing: since the regenerated Swift zip checksum changed while Package.swift still points at v0.3.3, we may need to bump the bindings/package version so SwiftPM does not fetch the existing v0.3.3 asset with the old checksum.

[ovitrif]

also LGTM 👍🏻 🎉