fix: Accept dtype_config instead of dtype_conversion in dtype_wrapper

The dtype_wrapper now expects dtype_config (the full dataclass) from OpenHCS
and extracts the default_dtype_conversion field from it. This fixes the
structural issue where dtype_config was being filtered before reaching
ArrayBridge, preventing dtype conversion from working.

Changes:
- Extract dtype_config from kwargs instead of dtype_conversion
- Get conversion mode from dtype_config.default_dtype_conversion
- Maintains backward compatibility with None fallback

Author: trissim

src/arraybridge/decorators.py

@@ -168,9 +168,12 @@ def _create_dtype_wrapper(func, mem_type: MemoryType, func_name: str):
 
     @functools.wraps(func)
     def dtype_wrapper(image, *args, slice_by_slice: bool = False, **kwargs):
-        # Get dtype_conversion from kwargs (injected by OpenHCS dtype_config)
-        dtype_conversion = kwargs.pop('dtype_conversion', None)
-        if dtype_conversion is None:
+        # Get dtype_config from kwargs (injected by OpenHCS)
+        dtype_config = kwargs.pop('dtype_config', None)
+        if dtype_config is not None:
+            # Extract the conversion mode from the config dataclass
+            dtype_conversion = dtype_config.default_dtype_conversion
+        else:
             dtype_conversion = DtypeConversion.PRESERVE_INPUT
 
         # Store original dtype