chore(deps): migrate to curve25519-dalek 5.0, rand 0.10, sha3 0.12

[dobby-coder]

What

Lands the RustCrypto + rand dependency stack from #15 that was previously blocked on curve25519-dalek shipping a release compatible with rand_core 0.10 / digest 0.11.

curve25519-dalek 5.0.0-rc.1 moves its public API (Scalar::random, Scalar::from_hash) to rand_core 0.10 (CryptoRng) and digest 0.11. Because those APIs flow through gg.rs's setup/keygen/sign and the h_helper SHA3-512 path, the whole group must move together — which is now possible.

Changes

Dependency	From	To	Notes
curve25519-dalek	4.1	=5.0.0-rc.1	pinned exactly (pre-release)
rand_core	0.6	0.10	CryptoRng now implies Rng; dropped the deprecated RngCore from generic bounds
sha3	0.10	0.12	uses digest 0.11
shake	—	0.1	new dep — Shake128 was moved out of sha3 into the standalone shake crate in 0.12
rand (dev)	0.8	0.10	OsRng removed → tests/bench/doctest use rand::rng() (ThreadRng); thread_rng() → rand::rng()

No public API change beyond the generic bound simplification (R: RngCore + CryptoRng → R: CryptoRng, which is equivalent since CryptoRng: Rng in rand_core 0.10).

⚠️ Pre-release caveat

curve25519-dalek 5.0.0-rc.1 is a release candidate, not a stable release (max_stable is still 4.1.3). It is pinned with = to block surprise rc bumps, mirroring the existing bincode-next =3.0.0-rc.13 pin. If the maintainer prefers to wait for curve25519-dalek 5.0 stable, this PR can sit until then — the migration itself is complete and green.

Verification

• cargo fmt --all -- --check ✅
• cargo test --release --all-features ✅ (15 tests + 1 doctest)
• cargo build --no-default-features ✅
• cargo build --target wasm32-unknown-unknown --all-features --lib ✅
• cargo build --target wasm32-wasip1 --all-features --lib ✅
• cargo bench --no-run ✅

No warnings. MSRV already declares 1.91.1 (bumped in the earlier bincode-next PR), which satisfies the new stack.

Refs #15

[dobby-coder]

Verdict: APPROVE (sign-off). Posted as a COMMENT-event review only because GitHub blocks a bot from formally approving its own PR — dobby-coder[bot] authored and reviewed this one. Treat this as a clean approval, not a withheld one.

Rules-Dobby-2 gatekeeper — cycle 1

Per-rule compliance check (all memory rules + repos/ibs/notes.md) and Review-Dobby-2's code review both come back clean: 0 actionable findings.

Verified:

• Migration matches repos/ibs/notes.md exactly: curve25519-dalek =5.0.0-rc.1, rand_core 0.10, sha3 0.12, new shake 0.1 (Shake128 moved out of sha3 in 0.12), dev rand 0.10. The bound change R: RngCore + CryptoRng → R: CryptoRng is equivalent since CryptoRng: Rng in rand_core 0.10. OsRng / thread_rng() → rand::rng() throughout tests/bench/doctest.
• PR body's verification matrix is consistent with the diff (fmt check, --all-features tests + doctest, --no-default-features build, wasm unknown + wasip1, bench --no-run).
• Refs #15 (not Closes #15) is the correct keyword here: #15 is a multi-dependency tracker and this PR lands only the RustCrypto/rand subset — criterion, csv, sha2, rand_xorshift bumps remain open. Tick the #15 checkboxes on merge rather than auto-closing the tracker.

One decision for a human: curve25519-dalek 5.0.0-rc.1 is a release candidate (max_stable is still 4.1.3), pinned with =. Accept the rc pin now or wait for 5.0 stable — maintainer's call. The migration is complete and green either way; flipping to ready-for-review to surface that decision.