DataFrame constructor is inconsistent in handling of duplicates in columns

[TomAugspurger]

In [5]: pd.DataFrame(columns=['a', 'a', 'b'])
Out[5]:
Empty DataFrame
Columns: [a, b]
Index: []

In [6]: pd.DataFrame.from_records({}, columns=['a', 'a', 'b'])
Out[6]:
Empty DataFrame
Columns: [a, a, b]
Index: []

I don't know which is preferred.

This is not being fixed in #24387

[TomAugspurger]

Whoops, my Out[5] was on the wrong branch. On master it matches from_records, which makes my life in #24387 harder.

In [9]: pd.DataFrame(columns=['a', 'a', 'b'])
Out[9]:
Empty DataFrame
Columns: [a, a, b]
Index: []