Fix: Enforce default embedding model in create_dataset / update_dataset (#8486)

### What problem does this PR solve?

Previous:
- Defaulted to hardcoded model 'BAAI/bge-large-zh-v1.5@BAAI'
- Did not respect user-configured default embedding_model

Now:
- Correctly prioritizes user-configured default embedding_model

Other:
- Make embedding_model optional in CreateDatasetReq with proper None
handling
- Add default embedding model fallback in dataset update when empty
- Enhance validation utils to handle None values and string
normalization
- Update SDK default embedding model to None to match API changes
- Adjust related test cases to reflect new validation rules

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

api/apps/sdk/dataset.py

@@ -347,6 +347,8 @@ def update(tenant_id, dataset_id):
                 return get_error_data_result(message=f"Dataset name '{req['name']}' already exists")
 
         if "embd_id" in req:
+            if not req["embd_id"]:
+                req["embd_id"] = kb.embd_id
             if kb.chunk_num != 0 and req["embd_id"] != kb.embd_id:
                 return get_error_data_result(message=f"When chunk_num ({kb.chunk_num}) > 0, embedding_model must remain {kb.embd_id}")
             ok, err = verify_embedding_availability(req["embd_id"], tenant_id)