refactor(account-keychain): use enumerable maps for scopes

[legion2002]

What This PR Does

This PR now contains only the enumerable-map refactor that remained after the TIP-1011 stack merged.

• adds EnumerableMap<K, V> to the precompile storage types, backed by a Set<K> for authoritative key membership and a Mapping<K, V> for values
• exports EnumerableMap from crates/precompiles/src/storage/types/mod.rs
• changes AccountKeychain scope storage so KeyScope.targets becomes EnumerableMap<Address, TargetScope> and TargetScope.selectors becomes EnumerableMap<FixedBytes<4>, SelectorScope>
• updates the AccountKeychain read and mutation paths to use the enumerable-map API instead of parallel Set + Mapping structures when enumerating, inserting, removing, and clearing target and selector scopes
• preserves the existing TIP-1011 behavior: is_scoped still distinguishes unrestricted vs scoped deny-all, and empty child sets still mean "no further restriction"

This does not change the external TIP-1011 interface or validation model. It only changes how scoped target/selector state is stored and maintained internally.

Verification

• cargo fmt --all --check
• cargo test -p tempo-precompiles --lib

[github-actions]

📊 Tempo Precompiles Coverage

📦 Download full HTML report

[tempoxyz-bot]

👁️ Cyclops Review — No actionable findings.

Nine independent audit passes (3 workers × 3 iterations) confirmed that the Set + Mapping → EnumerableMap refactor is storage-layout compatible, preserves all TIP-1011 invariants, and remains unreachable on the current T1C production hardfork.

Reviewer Callouts

• ⚡ Future EnumerableMap reuse with nested values: EnumerableMap::load()/store() intentionally return Fatal (enumerable_map.rs:L201-L210). Current code only uses handler-style access (keys(), contains(), remove(), clear(), set_with()). Re-check if whole-value at()/set() is ever used with nested struct values.
• ⚡ Keychain atomicity under T1C: AccountKeychain performs multi-step scope writes without a storage.checkpoint() guard. Pre-existing pattern, not introduced here, but worth revisiting if keychain mutators grow.
• ⚡ T3 activation re-review: All TIP-1011 call-scope entrypoints remain T3-gated. Run upgrade-compatibility tests when T3 activates on non-dev networks.

[legion2002]

we've chosen not to do this, because it does not have perf improvements over the enumerable set approach we have right now. We might reopen later, just to add an enumerable map impl to the PR