BUG: numpy_nullable NaNs do not round-trip through CSV

[jbrockmendel]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

arr = pd.array([0, pd.NA])
ser = pd.Series(arr) / 0
ser[2] = 3
df = ser.to_frame("A")

text = df.to_csv()

rt = pd.read_csv(StringIO(text), dtype_backend="numpy_nullable")[["A"]]

>>> df.loc[0, "A"]
np.float64(nan)
>>> rt.loc[0, "A"]
<NA>

Issue Description

This is a consequence of the constructor casting NaNs to pd.NA, xref #32265

Expected Behavior

NA

Installed Versions

Replace this line with the output of pd.show_versions()