Trim SGLang launch command to the verified minimal recipe

Trim the SGLang launch command to the verified minimal recipe.

All removed flags were verified as unnecessary on the current `lmsysorg/sglang:dev-Ling-3.0-flash-VL` image: serving smoke plus full MMMU-Pro / GSM8K / bench_serving were measured without them. The MXFP4 MoE backend and the ling3 parsers are auto-selected; `--mem-fraction-static` uses the engine default with no measured perf/accuracy impact.

See the cookbook for the per-hardware verified matrix: https://docs.sglang.io/cookbook/autoregressive/InclusionAI/Ling-3.0-flash-VL

README.md

@@ -70,7 +70,6 @@ docker run --rm --gpus all --ipc=host --shm-size 32g \
     --tp 4 \
     --context-length 262144 \
     --json-model-override-args '{"rope_scaling":{"rope_type":"yarn","factor":2.0,"rope_theta":6000000,"partial_rotary_factor":0.5,"original_max_position_embeddings":131072}}' \
-    --mem-fraction-static 0.85 \
     --trust-remote-code \
     --reasoning-parser auto \
     --tool-call-parser auto \