fix: Restore embedding-mode parameter to all examples

- All examples now have --embedding-mode parameter (unified interface benefit)
- Default is 'sentence-transformers' (consistent with original behavior)
- Users can now use OpenAI or MLX embeddings with any data source
- Maintains functional equivalence with original scripts

examples/document_rag.py

@@ -20,7 +20,7 @@ class DocumentRAG(BaseRAGExample):
         super().__init__(
             name="Document",
             description="Process and query documents (PDF, TXT, MD, etc.) with LEANN",
-            default_index_name="test_doc_files",  # Match original main_cli_example.py default
+            default_index_name="test_doc_files",
         )
 
     def _add_specific_arguments(self, parser):