GH-44188: [Python] Fix pandas roundtrip with bytes column names (#44171)

### Rationale for this change

There is a bug that when column dtype is np.bytes，it will goto the final branch and run level=level.astype(dtype)

### Are these changes tested?
Yes
* GitHub Issue: #44188

Lead-authored-by: Piong1997 <[email protected]>
Co-authored-by: Antoine Pitrou <[email protected]>
Signed-off-by: Antoine Pitrou <[email protected]>

Author: Piong1997

python/pyarrow/pandas_compat.py

@@ -1163,7 +1163,7 @@ def _reconstruct_columns_from_metadata(columns, column_indexes):
         if dtype == np.bytes_:
             level = level.map(encoder)
         # ARROW-13756: if index is timezone aware DataTimeIndex
-        if pandas_dtype == "datetimetz":
+        elif pandas_dtype == "datetimetz":
             tz = pa.lib.string_to_tzinfo(
                 column_indexes[0]['metadata']['timezone'])
             level = pd.to_datetime(level, utc=True).tz_convert(tz)

python/pyarrow/tests/test_pandas.py

@@ -5255,6 +5255,13 @@ def roundtrip(df, schema=None):
               schema=schema)
 
 
+def test_bytes_column_name_to_pandas():
+    df = pd.DataFrame([[0.1, 0.2], [0.3, 0.4]], columns=[b'col1', b'col2'])
+    table = pa.Table.from_pandas(df)
+    assert table.column_names == ['col1', 'col2']
+    assert table.to_pandas().equals(df)
+
+
 @pytest.mark.processes
 def test_is_data_frame_race_condition():
     # See https://github.com/apache/arrow/issues/39313