Inject contract deployments

[jochem-brouwer]

What this adds on top of PR 156

PR 156 generates the stateful benchmark payloads: gas-bump → funding → a
deploy phase (test_deploy_existing_contracts) → the measured benchmark run.
This branch extends that with:

1. Deploy output becomes prestate. The deploy phase's payloads are promoted
   into setup-global-test.txt (the base replayed before every test, like
   gas-bump.txt / funding.txt) instead of being a measured scenario, and the
   addon anchors benchmark scenarios on that deploy head. So predeployed
   contracts/EOAs exist before the measured run and the client under test does
   not deploy them during it (and therefore can't warm/cache them).

2. Two-fork runs (--benchmark-fork). Fill the setup/prestate on --fork
   and the benchmarks on a later fork. The mitm addon is restarted for the
   benchmark phase with a min_block_timestamp floor (read from the genesis
   fork-transition timestamp) so benchmark blocks land past the transition (and
   use the correct engine_newPayload version), while reuse_globals preserves
   the prestate across the restart. Omit --benchmark-fork → single-fork,
   unchanged behavior.

Paired execution-specs changes (as changes to PR 2947 (note: PR not to execution-specs but as PR on top of PR 2947 as PR 9))

• test_deploy_existing_contracts also creates EIP-7702 delegate EOAs
  (delegate i → i-th EXISTING_CONTRACT_DIFF contract) as part of the
  prestate.
• helpers.py: derivable diff_delegate_authority(i) /
  diff_delegate_target(fork, i) so any benchmark can target the delegates.
• New benchmark case diff_to_delegated_contract_diff that sends transfers to
  those prestate delegates.

How to use

1. Check out: gas-benchmarks inject-contract-deployments, execution-specs
   deployments.
2. Provide a genesis scheduling the benchmark fork at a high timestamp T
   (e.g. Osaka@0, Amsterdam@T via eip7928TransitionTimestamp and the other
   Amsterdam EIP transitions; the generator reads T from the genesis).
3. Run:

python3 eest_stateful_generator.py \
  --fork Osaka --benchmark-fork Amsterdam \
  --genesis-path <osaka-amsterdam-transition>.json \
  --chain mainnet --rpc-chain-id 1 \
  --rpc-seed-key <key> --rpc-address <addr> \
  --test-path tests/benchmark/stateful/bloatnet/ \
  --parameter_filter diff_to_delegated_contract_diff \
  --eest_mode repricing \
  --gas-bump-count <N> --gas-benchmark-values <V> \
  --eest-no-pull

Flow: boot Nethermind on the transition genesis → gas-bump + funding (Osaka) →
deploy phase (Osaka): contracts + 7702 delegates promoted to the
setup-global-test.txt prestate → addon restarts for Amsterdam → benchmark
(Amsterdam): transfers to the delegated EOAs.

Single-fork: drop --benchmark-fork and use a normal (non-transition)
genesis; the run behaves like PR 156 plus the prestate-promotion change.