Title: fix(factory): atomically deploy and initialize pools to close admin-hijack front-running window

[Cyber-Mitch]

Closes #79

create_pool now deploys and initializes the pool atomically — deploy_v2 and a call to the pool's initialize happen inside the same invocation, so there's no externally-observable window where a deployed pool exists but is uninitialized for an attacker to race.

Why this uses env.invoke_contract with raw args, not a typed FarmingPoolClient

An earlier version of this fix used FarmingPoolClient::new(&env, &pool_address).initialize(...). That fails to link: farming-pool's #[contractimpl] macro exports admin and transfer_admin as top-level contract functions, and linking that crate into factory's own cdylib (which exports functions of the same names) collides at link time:

rust-lld: error: duplicate symbol: admin
rust-lld: error: duplicate symbol: transfer_admin

cargo build/cargo test never catch this — they only compile the rlib crate type, never cdylib. The actual deployable target has to be built explicitly to catch it:

cargo build -p factory --target wasm32v1-none --release

This PR builds initialize's call via env.invoke_contract with raw Val arguments instead, avoiding the typed client and the symbol collision entirely. Verified by running the real WASM build target, not just cargo build/cargo test — confirmed clean both before and after this change.

Since env.invoke_contract doesn't need FarmingPoolClient, farming-pool moved back to [dev-dependencies] in factory/Cargo.toml (used only by tests, not at runtime) — this closes the collision risk for good, not just for admin/transfer_admin specifically.

Design decisions

• Admin: the factory's own admin (load_admin), not a caller-supplied delegate. Closes the front-running window, which is this issue's scope; per-pool admin delegation is #80's separate concern, not addressed here.
• stake_token reconciliation: asset is the same variable passed to both PoolRecord.asset and the pool's initialize call — they cannot diverge, closing the stake_token-hijack variant the issue also flags.
• Deliberately did not adopt #80-scoped additions from an existing unmerged commit that solves this same collision (a new global_multiplier parameter, daily_rate→credit_rate conversion, Result return type) — those are out of scope for #79.

Tests

• test_create_pool_pool_admin_is_set_atomically — admin() on the deployed pool equals the intended admin immediately after create_pool returns.
• test_third_party_cannot_reinitialize_deployed_pool — an unrelated address's initialize attempt on the deployed pool returns PoolError::AlreadyInitialized as a typed error, not a panic; confirms admin is unchanged.
• Both use #87's real-WASM integration test infrastructure.

Verification

Check	Result
cargo build --workspace	✅
cargo build -p factory --target wasm32v1-none --release (the real gate)	✅ links cleanly, no duplicate symbols
cargo test --workspace	✅ factory: 29 unit + 5 integration; farming-pool: 74 unit, all pass
cargo clippy -p factory --all-targets	✅ no new warnings

Cross-reference

Same root cause as #78 — resolving together as the issue requests.

Task completed
@prodbycorne Please Review

[netlify]

✅ Deploy Preview for sdcontracts ready!

Name	Link
🔨 Latest commit	60287a6
🔍 Latest deploy log	https://app.netlify.com/projects/sdcontracts/deploys/6a60ab0dc14bc40008cc1dbc
😎 Deploy Preview	https://deploy-preview-105--sdcontracts.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[Cyber-Mitch]

@prodbycorne Please Review