test: #698 — behavioral parity tests for argon2 + ethers integrations

[TheHypnoo]

Summary

Follow-up to #694, addressing #698. Converts deterministic entries from test_ffi_surface_stdlib_integrations.ts into focused behavioral parity fixtures.

Changes

• test_parity_argon2.ts + expected output — round-trip argon2.hash / argon2.verify with random-salt shape assertions. Only the async path is wired in Perry's NativeModSig table, so this covers the two FFI entries that are reachable end-to-end (js_argon2_hash, js_argon2_verify).
• test_parity_ethers.ts + expected output — deterministic numeric/address helpers: getAddress (EIP-55 checksum), parseEther/formatEther, parseUnits/formatUnits at 6- and 9-decimal positions. Covers the five js_ethers_* helpers that are pure functions of their inputs.
• test_parity_crypto.ts — adds a @covers block declaring the crypto/webcrypto/crypto_e2e FFI surface it already exercises via node:crypto and crypto.subtle (digest, hash, pbkdf2/hkdf, subtle.sign/verify).
• Regenerated test-coverage/COVERAGE.md and test_ffi_surface_* inventory files.

Both new fixtures use the test-parity/expected/ mechanism: Perry routes the npm-style imports to its bundled perry-ext-* wrappers, but Node can't load the same names without node_modules, so they fall through to the stored expected-output comparison.

Scope notes

I drafted fixtures for jsonwebtoken, bcrypt, and cheerio as well, but each exposed pre-existing bugs in the dispatch layer that are out of scope here:

• jsonwebtoken.sign — NativeModSig passes payload / secret as NA_F64 (NaN-boxed) but js_jwt_sign reads them as *const StringHeader, producing a garbage token (calling-convention mismatch — same shape as the argon2.hash(password) receives null StringHeader at the FFI boundary #591 argon2 fix that landed earlier).
• bcrypt.hash — return value is reported as typeof === "object" in user code (compare round-trips correctly, but .startsWith(...) on the hash fails).
• cheerio — $.select(...) returns a bare-number handle, so .length() / .first() etc. fail with (number).method is not a function at the prototype check.

These should each get their own follow-up issue. For now they stay listed in test_ffi_surface_stdlib_integrations.ts. Service-backed integrations (MongoDB, MySQL, PG, ioredis, Nodemailer, Sharp) also remain in the inventory pending the local-container / opt-in test setup described in #698.

Inventory delta

• test_ffi_surface_stdlib_integrations.ts: 156 → 138 unique FFI names (-18).
• ./test-coverage/audit.sh --markdown: still 100% TS FFI surface coverage.

Test plan

• ./test-coverage/regen_ts_surface_inventory.py
• ./test-coverage/audit.sh --markdown reports 100% TS coverage
• ./run_parity_tests.sh --filter test_parity_argon2 → PASS (expected-output)
• ./run_parity_tests.sh --filter test_parity_ethers → PASS (expected-output)
• ./run_parity_tests.sh --filter test_ffi_surface → 5/5 PASS

Checklist

• I have NOT bumped the workspace version or edited CLAUDE.md / CHANGELOG.md
• My commit follows the loose conventional prefix style used in the log
• I've read CONTRIBUTING.md and agree to the Code of Conduct