alloy-op-evm: reuse pool estimated da size and cache da scalar in OpBlockExecutor

With jovian active, `OpBlockExecutor::execute_transaction_without_commit` calls `jovian_da_footprint_estimation` for every transaction.

https://github.com/ethereum-optimism/optimism/blob/2258ea57dfad93ed72e7aa250bdcc00eb80dff64/rust/alloy-op-evm/src/block/mod.rs#L927

This is expensive as each call do:
- full fastLZ pass over the tx bytes via `op_revm::estimate_tx_compressed_size`
- 2 storage reads: `basic(L1_BLOCK_CONTRACT)` and `L1BlockInfo::fetch_da_footprint_gas_scalar`

https://github.com/ethereum-optimism/optimism/blob/2258ea57dfad93ed72e7aa250bdcc00eb80dff64/rust/alloy-op-evm/src/block/mod.rs#L507-L525

op-reth pool already lazily compute compressed size which can be used by the builder with `PoolTransaction::estimated_da_size()`

https://github.com/ethereum-optimism/optimism/blob/2258ea57dfad93ed72e7aa250bdcc00eb80dff64/rust/op-reth/crates/txpool/src/transaction.rs#L45

OpBlockExecutor could accept a caller supplied precomputed compressed size, e.g. alongside `WithEncoded` tx used if given, else fallback to compute it. Builder's implementors can then reuse pre-computation done in the pool.

Similarly, scalar is block-constant and can be fetched once and cached per block.

Initial benchmark saw a ~20% execution time improvement over a set of 200 simple txs.

Start in rust/alloy-op-evm/src/block/mod.rs at execute_transaction_without_commit and jovian_da_footprint_estimation, then read rust/op-reth/crates/txpool/src/transaction.rs for PoolTransaction::estimated_da_size(). Verify how WithEncoded transactions are passed and where the scalar is fetched. Done means reusing a supplied compressed-size estimate when available, falling back otherwise, and caching the block-constant scalar without changing execution results.

Written by the indexing model from the issue text.