Integrate with MoonEP and MXFP4 experts on the DeepGEMM runner

[lyppg]

MoonEP: MXFP4 experts on the DeepGEMM runner

Builds on #33249, this PR adds initial support for MoonEP a2a backend and DeepGemm runner for MXFP4 experts K3.

Validation

Validated end to end on Kimi-K3, 4x GB200, TP16/EP16, MXFP4 weights.
gsm8k test:

python -m sglang.test.run_eval --eval-name gsm8k --port 30000 --num-examples 500 --num-shots 5  --max-tokens 2048
Total latency: 643.858 s
Score: 0.986
Output throughput: 157.187 token/s
[METRIC] gsm8k_score=0.986 labels={"model": "Kimi-K3", "eval": "gsm8k"}
[METRIC] gsm8k_latency=643.8576721129939 labels={"model": "Kimi-K3", "eval": "gsm8k"}

---

CI States

Latest PR Test (Base): ❌ Run #31830931356
Latest PR Test (Extra): ❌ Run #31830931201

[lyppg]

moved to sgl-project#34874.