PR-21 test(worker): worker benchmark and tradeoff report to release v2.1.0 #46

Objective

Measure, don't assume — document whether Worker simulation is actually worth it for this workload.

Measure

• Main-thread frame p95
• Simulation p95
• Worker message latency
• Snapshot serialization cost
• Message count/sec

Note: for a small Snake simulation, Worker can cost more than it saves — that is an acceptable and technically valuable conclusion. The goal is learning and evidence, not automatically claiming Worker is faster.

Exit criteria (closes Phase 5 / M5)

• Benchmark documents tradeoffs
• Tag/release as v2.1.0