refactor: extract _process_tensors_microscale to reduce duplication

Signed-off-by: David Zheng <dqzheng1996@gmail.com>

Author: dzhengAP

src/llmcompressor/entrypoints/model_free/process.py

@@ -177,9 +177,9 @@ def process_file_group_microscale_scheme(
         "Use `process_file` or `process_file_microscale_scheme` for "
         "non-microscale schemes"
     )
-    assert len(file_paths) == len(save_paths), (
-        "file_paths and save_paths must have the same length"
-    )
+    assert len(file_paths) == len(
+        save_paths
+    ), "file_paths and save_paths must have the same length"
 
     # Load all tensors from the group, tracking which output shard each belongs to
     tensor_to_shard: dict[str, str] = {}
@@ -308,4 +308,4 @@ def _process_tensors_microscale(
                 if tensor_to_shard is not None:
                     tensor_to_shard[key] = original_shard
 
-    return tensors, tensor_to_shard
+    return tensors, tensor_to_shard