FP8 updated: Fix ComfyUI quantization metadata

Added missing comfy_quant tensors, _quantization_metadata header, and scaled_fp8 marker so ComfyUI recognizes the model as scaled FP8 (MixedPrecisionOps), resolving the blank/black output issue.

qwen-image-2.1-UC-fp8.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6ff3cb225b7dfeadbfdbae78daee0b126f5dd2481b94ae0a48c7a98de4b15fc3
-size 7116769044
+oid sha256:877b77290339986bca8f771aead381c7a5777347c27d252bbaa915f76667ab0b
+size 7116818183