REGR: DatetimeTZDtype.__from_arrow__ interprets UTC values as local wall time

[jorisvandenbossche]

Using the latest pandas:

In [6]: ser = pd.Series(pd.to_datetime(["2012-01-01", "2012-01-02"]).tz_localize("Europe/Brussels"))

In [7]: ser
Out[7]: 
0   2012-01-01 00:00:00+01:00
1   2012-01-02 00:00:00+01:00
dtype: datetime64[ns, Europe/Brussels]

In [8]: pa.array(ser)
Out[8]: 
<pyarrow.lib.TimestampArray object at 0x7f695b4625c0>
[
  2011-12-31 23:00:00.000000000,   # <-- those values show the UTC timestamps, so this is correct
  2012-01-01 23:00:00.000000000
]

In [9]: ser.dtype.__from_arrow__(pa.array(ser))
Out[9]: 
<DatetimeArray>
['2011-12-31 23:00:00+01:00', '2012-01-01 23:00:00+01:00']   # <-- still the UTC wall time, but with tz offset -> wrong
Length: 2, dtype: datetime64[ns, Europe/Brussels]

I assume this is caused by #56043 cc @jbrockmendel

[jbrockmendel]

In DatetimeTZDtype.__from_arrow__, the following edit seems to fix this:

+        return DatetimeArray._simple_new(np_arr, dtype=self)
-        return DatetimeArray._from_sequence(np_arr, dtype=self, copy=False)