Stop the engine from supplying a model id the remover must reject

InvisibleEngine substituted DEFAULT_MODEL_ID whenever model_id was None. When
b0ca205 tightened the remover's fixed-stack check from `not in {None,
DEFAULT_MODEL_ID}` to `is not None`, that substitution turned every single
InvisibleEngine construction into a ValueError - including the deployed Modal
worker's setup(), which is how it was found.

The library suite missed it because these two are tested from opposite sides:
every remover test builds WatermarkRemover directly with model_id unset, and
every engine test mocks the remover away. Nothing exercised the seam between
them. TestEngineDoesNotFabricateAModelId now does, without a GPU.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/remove_ai_watermarks/invisible_engine.py

@@ -121,10 +121,11 @@ class InvisibleEngine:
 
         from remove_ai_watermarks._internal.watermark_remover import WatermarkRemover
 
-        effective_model = model_id or self.DEFAULT_MODEL_ID
-
+        # Pass model_id through untouched. Substituting DEFAULT_MODEL_ID for None here
+        # meant the engine always supplied a model the remover is required to reject,
+        # so EVERY construction raised once that check tightened to "is not None".
         self._remover = WatermarkRemover(
-            model_id=effective_model,
+            model_id=model_id,
             device=device,
             progress_callback=progress_callback,
             hf_token=hf_token,