Add 9 blockchain tests (28 parametrized items) targeting inter-client
disagreements observed on BAL devnet-3 where EL clients diverged on
block.gas_used due to incorrect 2D gas metering.

Tests cover:
- State-dominant blocks (SSTORE state gas > regular gas)
- Regular-dominant blocks (STOP txs, zero state gas)
- Mixed tx blocks (grouped and interleaved ordering)
- EIP-7778 interaction (refunds excluded from block gas)
- Boundary: block valid when max(regular, state) == gas_limit
- CALL creating new account (GAS_NEW_ACCOUNT state gas)
- CREATE tx intrinsic state gas
- Multi-block dimension flip (regular→state dominance)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Use fork methods for all gas computation: Op.SSTORE.gas_cost(fork),
Op.SSTORE.with_metadata() for warm slot dynamics, fork.sstore_state_gas(),
fork.transaction_intrinsic_cost_calculator(). No manual gas constants.

Rename helpers: drop _ prefix (sstore_tx_gas, sstore_txs, stop_txs).
Add return type annotations. Remove _stop_tx_gas wrapper. Use
stop_txs() helper consistently. Replace hardcoded 0xDEAD0001 with
pre.fund_eoa(amount=0). Replace hardcoded +1000 padding with
fork-derived value. Remove unused Bytecode import and explicit fee
params. Clean up docstrings and test ids.