BUG/API: constructing nullable integer and floating arrays from boolean data

[mzeitlin11]

From #41970 (comment), right now constructing nullable integer (and floating) arrays from boolean data has the following behavior:

In [2]: pd.Series([True, False, True, True], dtype='Int64')
Out[2]:
0    1
1    0
2    1
3    1
dtype: Int64

In [3]: pd.Series([True, False, True, pd.NA], dtype='Int64')
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
...
TypeError: object cannot be converted to an IntegerDtype

This seems inconsistent, if so, which behavior is correct?

[mzeitlin11]

I think having the second case not raise makes sense from a consistency perspective given that you can do something like

>>> pd.Series([True, np.nan, True], dtype=float)
0    1.0
1    NaN
2    1.0