[BUG] cudf.pandas groupby-agg does not match pandas when column names are boolean literals

[Matt711]

Describe the bug
See reproducer

Steps/Code to reproduce bug

In [1]: import pandas as pd

In [2]: pd.DataFrame({True: [1, 2], False: [3, 4]}).groupby(True, sort=False)[False].max()
Out[2]: 
True
1    3
2    4
Name: False, dtype: int64

In [1]: %load_ext cudf.pandas

In [2]: import pandas as pd

In [3]: pd.DataFrame({True: [1, 2], False: [3, 4]}).groupby(True, sort=False)[False].max()
Out[3]: 
Empty DataFrame
Columns: []
Index: [1, 2]

Expected behavior
match pandas

[Matt711]

Found during the Narwhals bump to 1.47 #19358