[misc]: DreamX — point registry/examples at published Diffusers weights

Verified on GB200: both FastVideo/DreamX-World-5B-Cam-Diffusers and
FastVideo/DreamX-World-5B-Diffusers strict-load from the Hub with no
missing/unexpected keys, and the Cam example generates a video end-to-end
from the Hub repo id.

- registry: hf_model_paths now reference the published FastVideo Diffusers
  repos instead of the official GD-ML checkpoints (which are raw-format and
  not directly loadable).
- example + support matrix: default/documented model ids updated likewise.
- SSIM test: model paths default to the Hub repos like sibling SSIM tests;
  the local-converted-path skip machinery is gone (env overrides remain).

examples/inference/basic/basic_dreamx_world.py

@@ -15,7 +15,7 @@ def _env_float(name: str, default: float) -> float:
 
 
 def main():
-    model_name = os.getenv("DREAMX_WORLD_MODEL_DIR", "GD-ML/DreamX-World-5B-Cam")
+    model_name = os.getenv("DREAMX_WORLD_MODEL_DIR", "FastVideo/DreamX-World-5B-Cam-Diffusers")
     generator = VideoGenerator.from_pretrained(
         model_name,
         num_gpus=1,