feat: Implement ONNX model inference for BEN2, Depth Anything 3, and MoGe

- Added `ben2_onnx.py` for BEN2 ONNX inference with preprocessing.
- Introduced `da3_onnx.py` for Depth Anything 3 inference, including camera projection calculations.
- Created `moge_onnx.py` for MoGe-2 ONNX inference and post-processing.
- Implemented shared ONNX Runtime session helpers in `onnx_runtime.py`.
- Added model synchronization script `sync_models.py` for downloading and verifying ONNX models.
- Developed unit tests for BEN2, Depth Anything 3, and MoGe ONNX models to ensure functionality.
- Updated `vendor.lock.toml` to remove unused artifacts.

prepare_vendor.py

@@ -65,7 +65,7 @@ def download_artifact(artifact, destination_dir=VENDOR_DIR):
 
 
 def prepare(platforms=None, lock_path=LOCK_PATH, destination_dir=VENDOR_DIR):
-    """Prepare common third-party archives for selected platforms."""
+    """Prepare bundled uv executables for selected platforms."""
     lock = load_lock(lock_path)
     available = lock["platforms"]
     selected = list(available) if platforms is None else platforms
@@ -73,8 +73,7 @@ def prepare(platforms=None, lock_path=LOCK_PATH, destination_dir=VENDOR_DIR):
     if unknown:
         raise ValueError(f"Unsupported platform: {', '.join(unknown)}")
 
-    artifacts = list(lock["artifacts"].values())
-    artifacts.extend(available[platform] for platform in selected)
+    artifacts = [available[platform] for platform in selected]
     return [
         download_artifact(artifact, destination_dir)
         for artifact in artifacts