feat: warmup_test_payload phase between setup and test

[skylenet]

Summary

Adds an optional warmup phase that runs between setup and test. For each engine_newPayload* call in the test step, the runner sends a modified copy with blockHash recomputed so the EL still does real payload decoding, transaction validation, and (depending on the chosen method) state-root execution before rejecting on the mismatch. The value is in that work — cache fills and codepath warming — not in the eventual rejection.

Two mutation strategies are supported, selected via method:

• invalid-stateroot (default): replaces stateRoot with a deterministic placeholder derived from a fork-specific salt and the iteration index, then recomputes blockHash. EL typically rejects on state-root mismatch after executing the block.
• invalid-gasused: keeps stateRoot intact and subtracts (1+i) from gasUsed for iteration i (so iteration 0 = original-1, iteration 1 = original-2, …), then recomputes blockHash. EL typically rejects once it notices the gas mismatch.

count controls how many times each engine_newPayload* line is sent. Each iteration produces a distinct payload (per-iteration salted stateRoot, or per-iteration gasUsed decrement) so the client treats the calls as distinct. Non-newPayload lines (e.g. engine_forkchoiceUpdated*) are sent once regardless.

Configured via:

runner:
  client:
    config:
      warmup_test_payload:
        enabled: true
        fork: osaka
        method: invalid-stateroot   # or invalid-gasused; defaults to invalid-stateroot
        count: 3                    # optional; defaults to 1

Currently only fork: osaka is supported. Standard global/instance override pattern applies — instance-level config (when non-nil) fully replaces the global default.

• New pkg/warmup package: rewrites engine_newPayload* JSON-RPC lines via a method-dispatched mutator, recomputes blockHash using go-ethereum's engine.ExecutableDataToBlockNoHash. Non-newPayload lines pass through unchanged. invalid-gasused returns an error if count would underflow the original gasUsed.
• Config: WarmupTestPayloadConfig (with Method, Count, Fork, Enabled), GetWarmupTestPayload(instance) getter, validator that requires fork == "osaka", accepts either method, and rejects negative counts.
• Executor: new StepTypeWarmup, runs after setup / before test. Results land at <test>/warmup.{response,result-details.json,result-aggregated.json} and surface in result.json and per-suite stats. Warmup gas/time is recorded but not folded into the test totals so test metrics stay comparable across warmup-on/off runs.
• Runner: warmup config plumbed into all three ExecuteOptions call sites (lifecycle + container + checkpoint) and the resolved value lands in the run's config.json under instance.warmup_test_payload.
• UI: rendered in the run-detail Configuration panel (shows method, fork, count) and as a row in the compare ConfigDiff (with "differs" highlighting).
• Docs: docs/configuration.md updated with a Warmup Test Payload section covering both methods + entries in the global and instance-override tables.

New direct dep: github.com/ethereum/go-ethereum v1.17.2.

Test plan

• go test ./... passes (incl. new pkg/warmup tests for both methods, count-produces-distinct-variants, underflow guard, and pkg/config tests for getter + validator)
• golangci-lint run --new-from-rev="origin/master" clean
• tsc -b and eslint clean for the UI changes
• End-to-end with method: invalid-stateroot, fork: osaka: confirm warmup payloads are sent and the client reaches state-root execution before rejecting (vs. rejecting at hash check)
• End-to-end with method: invalid-gasused, fork: osaka: confirm warmup payloads are sent and the client reaches gas-validation before rejecting; verify count=N produces N distinct payloads per newPayload
• Verify <test>/warmup.* files are written and result.json carries steps.warmup
• Verify the run-detail page shows the Warmup Test Payload card with the correct method and count, and the compare page row reflects differences across runs