Preserve Date/Datetime index name in yf.download() output

reindex_dfs() rebuilt the combined index via sorted(set(...)) +
pd.to_datetime, dropping the index name set by history.py ('Date' for
daily, 'Datetime' for intraday). After concat, data.index.name was
None, so users doing data.stack().reset_index() got a 'level_0' column
instead of 'Date' / 'Datetime'.

Use Index.union, which preserves the name natively when sources agree.

Drive-by: narrow data.columns to MultiIndex before swaplevel to silence
a pyright attribute warning.

tests/test_ticker.py

@@ -324,6 +324,8 @@ class TestTickerHistory(unittest.TestCase):
                                                interval=interval, progress=False)
                             self.assertIsInstance(data, pd.DataFrame, "data has wrong type")
                             self.assertFalse(data.empty, "data is empty")
+                            expected_name = 'Datetime' if interval[-1] in ('m', 'h') else 'Date'
+                            self.assertEqual(data.index.name, expected_name)
                             if ignore_tz:
                                 self.assertIsNone(data.index.tz)
                             else:

yfinance/multi.py

@@ -214,7 +214,7 @@ def _download_impl(ctx, tickers, start=None, end=None, actions=False, threads=Tr
                           keys=ctx.dfs.keys(), names=['Ticker', 'Price'])
     data.rename(columns=ctx.isins, inplace=True)
 
-    if group_by == 'column':
+    if group_by == 'column' and isinstance(data.columns, _pd.MultiIndex):
         data.columns = data.columns.swaplevel(0, 1)
         data.sort_index(level=0, axis=1, inplace=True)
 
@@ -240,11 +240,12 @@ def reindex_dfs(dfs, ignore_tz):
                 if (dfs[tkr] is not None) and (not dfs[tkr].empty):
                     dfs[tkr].index = dfs[tkr].index.tz_convert(tz_mode)
 
-    all_indices = set()
+    idx = None
     for df in dfs.values():
-        all_indices.update(df.index)
-    idx = sorted(all_indices)
-    idx = _pd.to_datetime(idx)
+        if df is not None and not df.empty:
+            idx = df.index if idx is None else idx.union(df.index)
+    if idx is None:
+        idx = _pd.DatetimeIndex([])
     for key, df in dfs.items():
         dfs[key] = df.reindex(idx)