docs: make homepage result-first with free5GC and vLLM evidence - #9

Summary

Rework the Gthulhu homepage so the first screen leads with measured workload outcomes, not architecture narrative.

What changes

• add a result table at the top of the English and Chinese homepages;
• highlight the published free5GC GTP scheduling result:
  • average UE ping latency: 88.98 ms → 2.079 ms (97.66% lower)
  • maximum UE ping latency: 130.95 ms → 8.45 ms (93.55% lower)
• highlight the reproducible vLLM CPU-pressure benchmark from [Blog] Keeping vLLM Fast Under CPU Pressure: An sched_ext Scheduler for GPU Inference vllm-project/vllm-project.github.io#300:
  • baseline decode around 6–7 t/s under CPU contention
  • tg128 around 21.3 t/s with Gthulhu + tiered scheduling policy
• clearly mark the vLLM result as an upstream-reviewing community benchmark, not a generalized performance guarantee;
• move Claim2Core architecture/roadmap below the proof points instead of making it the first message visitors see.

Why

The current homepage explains the architecture well, but makes visitors understand the system before showing why they should care. This change follows the result-first style used by performance-oriented infrastructure projects and makes existing published/community evidence the primary entry point.

Sources linked directly from the homepage:

• free5GC: https://free5gc.org/blog/20251126/20251126/
• free5GC earlier case study: https://free5gc.org/blog/20250726/index.en/
• vLLM blog PR: [Blog] Keeping vLLM Fast Under CPU Pressure: An sched_ext Scheduler for GPU Inference vllm-project/vllm-project.github.io#300