BUG: DataFrame(dict_of_series) raises depending on order of dict

[jbrockmendel]

• 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 master branch of pandas.

Reproducible Example

dti = pd.date_range("2016-01-01", periods=3)
ser1 = pd.Series(range(3), index=dti)
ser2 = pd.Series(range(3), index=dti.tz_localize("UTC"))
ser3 = pd.Series(range(3))

>>> pd.DataFrame({"A": ser3, "B": ser2, "C": ser1})  # works fine

>>> pd.DataFrame({"A": ser1, "B": ser2, "C": ser3})
TypeError: Cannot join tz-naive with tz-aware DatetimeIndex

Issue Description

When we pass a dict of Series, we end up going through indexes.api.union_indexes, which checks if our first index is a DatetimeIndex and if so dispatches to DatetimeIndex.union_many, which in this case raises. If the DatetimeIndex is not first, then it goes through Index.union and casts to object.

Then there's the issue that we don't pass sort to union_many so it gets ignored.

Expected Behavior

Cast to object.

Installed Versions

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