refactor(tests): Remove fork.gas_costs() reconstruction from Amsterdam EIP-8037/8038 tests

[marioevz]

Description

Motivation

Many Amsterdam gas tests re-derived expected gas by hand-summing fork.gas_costs() constants (NEW_ACCOUNT, CALL_VALUE, COLD_STORAGE_WRITE, VERY_LOW, …). This duplicates the framework's own gas calculation and makes the tests brittle: any future fork that reprices or restructures the schedule silently diverges from these hand-built expectations.

This PR replaces that reconstruction with framework constructs that track the live schedule, and fixes the one place the framework couldn't express a cost.

Scope: the eip8037_state_creation_gas_cost_increase and eip8038_state_access_gas_cost_increase suites (24 test files) plus a small testing-framework change.

Types of change

Framework (packages/testing)

• Wire CALL/CALLCODE new-account gas into the EIP-8037 two-dimensional model: their NEW_ACCOUNT charge now lands on the state dimension (opcode_state_map + a regular-cost override) instead of being stuck in regular gas, so .state_cost(fork) / .regular_cost(fork) are correct for value-to-new-account calls. Total .gas_cost(fork) is unchanged.
• Add a call_value_stipend() fork accessor for the value-call stipend, which has no opcode/bytecode representation.

Tests — replace hand-reconstructed gas with framework constructs

• Opcode/bytecode cost methods (.gas_cost / .regular_cost / .state_cost / .refund) on the actual fragments under test, driven by opcode metadata (account_new, value_transfer, address_warm, storage-slot state, …).
• Transaction-level costs via transaction_intrinsic_cost_calculator, transaction_top_frame_state_gas, and transaction_data_floor_cost_calculator.
• Fork-transition / cross-fork comparisons derived from the same bytecode/intrinsic evaluated at each fork (fork.fork_at(...)), not raw constants.
• The value-call stipend sourced from the new call_value_stipend() accessor.

Tests — drop redundant self-checks

• Removed assertions that compared a framework-computed value against a fork.gas_costs() decomposition of the same fork (framework-vs-framework). They add no coverage over the runtime behavior each test already exercises and only break on repricing.

Bug fix

• Corrected several contract-creation tests that subtracted NEW_ACCOUNT from the transaction intrinsic; under EIP-2780 the intrinsic is regular-only, so the subtraction went negative (masked previously because the state dimension dominated the header gas_used).

Deliberately retained

• One fork.gas_costs() use remains in the transient-storage regression test, where the declared OPCODE_TLOAD/OPCODE_TSTORE schedule value is the subject under test (measured at runtime via CodeGasMeasure).

Testing

• Full Amsterdam fixture fill: 9285 passed, 9 skipped (unchanged from baseline — no regressions).
• Framework opcode-gas unit tests pass.

Related Issues or PRs

Fixes #3042

Checklist

• Ran fast static checks to avoid CI fails, see Code Standards & Verifying Changes: just static
• PR title has the form <type>(<area>): <title>, where <type> and <area> come from an appropriate C-<type>, respectively A-<area>, label. The title should match the target squash commit message.

Cute Animal Picture

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.39%. Comparing base (5fa5938) to head (2743747).
⚠️ Report is 2 commits behind head on forks/amsterdam.

Additional details and impacted files

@@               Coverage Diff                @@
##           forks/amsterdam    #3169   +/-   ##
================================================
  Coverage            93.39%   93.39%           
================================================
  Files                  624      624           
  Lines                37014    37014           
  Branches              3386     3386           
================================================
  Hits                 34570    34570           
  Misses                1671     1671           
  Partials               773      773           

Flag	Coverage Δ
unittests	93.39% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[marioevz]

Updated write-test.md claude skill to set the new test expectations as directives.

[LouisTsai-Csie]

Finish eip-8037 review except tests/amsterdam/eip8037_state_creation_gas_cost_increase/test_state_gas_selfdestruct.py

I will leave my review comment later

[LouisTsai-Csie]

EIP-8037: Two small issues

• Missing coverage: test_create_refund_credited_against_child_spill still uses gas_costs.NEW_ACCOUNT and isn't covered in this refactor.
• Redundant state-gas definition: In test_exact_coinbase_fee_simple_sstore, we define sstore_state_gas = Op.SSTORE(new_value=1).state_cost(fork), then define sstore_code = Op.SSTORE(sstore_storage.store_next(1), 1), and reference sstore_state_gas. We could simplify by using sstore_code.state_cost(fork) directly instead. No changes needed, pure suggestion

EIP-8038: I need additional review time since I'm uncertain why some assertions were completely removed.

[marioevz]

@LouisTsai-Csie applied your comments and repurposed the test that was no longer correct (It checked an account creation refund when we don't charge for it in the first place anymore).