Add test coverage for cas-lib 0.2.85 APIs (#90)

[WingZer0o]

Summary

Audit of issue #90 found that every API it lists (AES-GCM-SIV, ML-KEM-1024, SLH-DSA, PBKDF2, Argon2 AES key derivation, AES-GCM key_from_vec, Ed25519 verify-with-key-pair) is already exposed through the FFI and TS layers via #91 and #92, and nothing else public in cas-lib 0.2.85 is unexposed. This PR fills the remaining test gaps.

TS tests (Playwright)

• AES-GCM-SIV: X25519 shared-secret key derivation (128/256) agrees on both sides and round-trips; KeyFromBytes keys round-trip; key/nonce/tag sizes; tampered ciphertext and wrong nonce throw
• Ed25519: verifyWithKeyPairBytes rejects a different key pair and a tampered signature, agrees with public-key verify, throws on wrong-length inputs
• PBKDF2: iteration count and password change the key; derive generates a fresh salt per call; derived key works as an AES-256-GCM key
• SLH-DSA: wrong verification key / tampered signature fail; wrong-length signature throws
• ML-KEM-1024: fresh encapsulation per call; wrong secret key yields a different shared secret; shared secret keys AES-256-GCM
• New public-api.spec.ts: guards barrel exports of the new wrappers/methods

Rust tests

• AES-GCM-SIV key from X25519 shared secret, end to end
• Ed25519 key-pair verify: tampered message returns false, bad lengths error

Reviewer notes

• AES-GCM-SIV key_from_x25519_shared_secret runs input through HKDF with no length check, so a wrong-length shared secret does not throw. Strict validation would be a cas-lib change.
• The comment on Expose remaining cas-lib 0.2.85 APIs through the FFI layer #90 saying Argon2 derive_aes_* was deferred is stale — Argon2 derive AES-128/256 keys #92 exposed it (salt is still random and not returned, so keys are non-reproducible; tests assert that).
• No production code changed.

Test plan

• cargo test --release — 61 passed
• npx playwright test on the 6 affected spec files — 114 passed (all 3 projects)

Closes #90

🤖 Generated with Claude Code