### What this PR does / why we need it? Add quantization param for `deepseek-w8a8` multi-node test ### Does this PR introduce _any_ user-facing change? ### How was this patch tested? - vLLM version: v0.11.0rc3 - vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0 Signed-off-by: wangli <wangli858794774@gmail.com>