feat: add finalized reward allocation ledger

[olatundefay-prog]

Closes #370

V2-SC-019 — Implement Reward Allocation Ledger

[dDevAhmed]

@olatundefay-prog review of head 27757d7c1239b871c74e719020372c87368c96a6: assignment and dependencies are valid, but this protocol-critical rewards ledger is not mergeable yet.

• fund() records the requested amount without checking the allocator's balance delta. A fee-on-transfer or otherwise non-standard ERC-20 can make _funded and claimable entitlements exceed actual custody. Use exact received-amount accounting (or explicitly reject transfer mismatches) and add regression coverage.
• Funds can be stranded: fund() remains callable after settlementId is finalized; zero allocations are accepted; partial allocation leaves an unrecoverable remainder; and multiple assets may be funded under one settlement although _finalized[settlementId] permits finalizing only one asset. Define the canonical single-/multi-asset model and ensure every funded unit is allocated, explicitly retained/refunded, or recoverable without cross-settlement consumption.
• The allocator accepts caller-supplied outcome, categories, weights, and totals without enforcing outcome/category compatibility or binding them to the frozen final settlement record. Integrate the narrow canonical settlement/final-resolution interface so disputed, undisputed, conclusive, and inconclusive paths cannot produce contradictory allocations.
• supportsInterface() omits type(IV2Module).interfaceId, despite the V2 module contract requiring it. Add the explicit interface response and conformance test.
• The deployment/documentation is contradictory: it says to register the allocator under SETTLEMENT, while _onlySettlementModule() expects the registered SETTLEMENT implementation to call the allocator. Registering the allocator there would replace the settlement engine and make external funding/finalization impossible. Give the ledger its canonical module identity, preserve SETTLEMENT as the authorized caller, and add deployment/composition coverage.
• Required verification is missing: the PR has only three Hardhat tests and no complete conclusive true/false, inconclusive, disputed/undisputed, category-compatibility, rounding/overflow boundary, duplicate-category, insufficient-claimable, exact-custody, accounting-invariant fuzz, or settlement-integration coverage. Use Math.mulDiv or prove multiplication bounds for amount * weight.

The exact-head CI, Fuzz Testing, and Gas Regression Detection runs are action_required; this is maintainer workflow authorization, not a contributor failure. After the fixes, run npm run compile && npm test && npm run compile:forge && forge test -vv && npm run test:gas, then request authorized fresh Actions and an independent human maintainer approval on the exact SHA.