[MXFP8][XPU] enable mxfp8 using vLLM main repo method (#3782)

Signed-off-by: Chendi Xue <chendi.xue@intel.com>

.buildkite/scripts/hardware_ci/run-xpu-test.sh

@@ -52,6 +52,7 @@ time timeout -k 30 30m docker run \
     pip install tblib==3.1.0
     cd /workspace/vllm-omni
     pytest -v -s -m "core_model and xpu and B60"
+    pytest -v -s tests/diffusion/quantization/test_mxfp8_config.py
     export VLLM_XPU_USE_SAMPLER_KERNEL=0    # NOTE: Remove this after vLLM v0.21.1 is merged. Fixes Qwen2-5 omni-expansion tests.
     pytest -v -s -m "advanced_model and xpu and B60"
 '