feat(load-test): separate benchmark setup and measurement

[meyer9]

Summary

• start the native load tester during benchmark setup and wait until its mempool prefill is ready
• coordinate the measured window through the CLI-only ready/start/started/finished handshake added in feat(load-tests): open-loop load generation, adaptive pacing, and B20 payloads base#4156
• force unbounded saturation mode with target_gps null and default mempool_target_blocks to 3
• stop metric collection when measured generation finishes, then build unmeasured settlement blocks until the worker drains
• propagate setup, premature-exit, and settlement failures instead of reporting a successful benchmark

Dependency

Depends on base/base#4156.

Validation

• go test ./runner/payload/loadtest ./runner/payload/worker ./runner/network/...
• GOFLAGS=-buildvcs=false make build
• git diff --check

The same-host ZFS A/B benchmark moved the first measured block from 0.00038B gas to 1.632B gas and reduced per-block gas CV from 94.9% to 4.66%.

[cb-heimdall]

🟡 Heimdall Review Status

Requirement	Status	More Info
Reviews	🟡 0/1	Denominator calculation Show calculation 1 if user is bot 0 1 if user is external 0 2 if repo is sensitive 0 From .codeflow.yml 1 Additional review requirements Show calculation Max 0 0 From CODEOWNERS 0 Global minimum 0 Max 1 1 1 if commit is unverified 1 Sum 2