BUG: fix Frame getitem when column keys are nested tuples pandas-dev#43780

Multi-column selection when the column keys are nested tuples (for
example when a MultiIndex has at each level a tuple) raised error due
to a condition in asarray_tuplesafe.

A key list of form [((1,))] converts to [[[1]]] through the call to
np.array, resulting in ndim>2. However, only the case ndim=2 was
further handled, causing a multi-dimensional array to be returned and
subsequently raising error.

This commit extends the ndim=2 condition for these type of keys.

Author: Svanazar

pandas/core/common.py

@@ -240,7 +240,8 @@ def asarray_tuplesafe(values, dtype: NpDtype | None = None) -> np.ndarray:
     if issubclass(result.dtype.type, str):
         result = np.asarray(values, dtype=object)
 
-    if result.ndim == 2:
+    # For ndim > 2 distinguish between nested tuples and multidimensional arrays
+    if result.ndim == 2 or (result.ndim > 2 and isinstance(values[0][0], tuple)):
         # Avoid building an array of arrays:
         values = [tuple(x) for x in values]
         result = construct_1d_object_array_from_listlike(values)