[executor] wire batchExecute calldata through Simulator::simulate gate

[obchain]

Tracking issue spun out of #208.

Context

Batcher::encode_calldata returns Bytes that are not self-validating. Per CLAUDE.md, every liquidation tx must pass an eth_call simulation gate before broadcast. The single-opportunity path is covered by TxBuilder → Simulator::simulate; the batch path does not yet have an equivalent call site wired up in the CLI pipeline.

The Rustdoc on encode_calldata now documents this contract explicitly and requires callers to run the returned calldata through Simulator::simulate before broadcast.

Scope

• Add a batch-aware simulation helper that takes a LiquidationBatch + Bytes and dispatches Simulator::simulate against the CharonLiquidator contract.
• Wire it into the CLI pipeline once the batcher graduates from planner-only to a producer of submitted txs.
• Add a fork test that broadcasts a batch after a successful simulate and asserts BatchExecuted emission, mirroring the existing single-opp test.

Acceptance

• No code path outside tests calls encode_calldata without passing the result through Simulator::simulate first.
• Fork test covers simulate-then-broadcast for batchExecute.
• CLAUDE.md safety invariant (eth_call gate on every broadcast) remains unviolated for the batch path.

Refs #208.

[obchain]

Fixed on feat/20-multi-liq-batcher (225a68b): Batcher::encode_calldata now returns UnsimulatedBatchCalldata; the only path out of that wrapper is Batcher::simulate, which produces SimulatedBatchCalldata — the only type the (future) batch broadcaster will accept. The CLAUDE.md eth_call-gate invariant is now a compile-time guarantee for the batch path. Foundry-side batchExecute fork coverage tracked separately on feat/25 / PR #53.