soiling_hsu doesn't work with read_tmy3 - index must be monotonic

[mikofski]

Describe the bug
calling pvlib.losses.soiling_hsu(read_tmy3(<tmy file>, coerce_year=<year>), ...) raises ValueError: index must be monotonic because the last hour of a TMY3 file has been coerced to the same year as the first index, instead of the first hour of the next year

To Reproduce
Steps to reproduce the behavior:

from pvlib.losses import soiling_hsu
from pvlib.iotools import read_tmy3
from pvlib.tests.conftest import DATA_DIR

greensboro = read_tmy3(DATA_DIR / '723170TYA.CSV', coerce_year=1990)
greensboro_rain = greensboro[0].Lprecipdepth
soiling_hsu(greensboro_rain, cleaning_threshold=6, tilt=10, pm2_5=1e-5, pm10=1e-5)

raises ValueError: index must be monotonic

Details

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-95-ff36a88e4e3a> in <module>
----> 1 soiling_hsu(greensboro_rain, cleaning_threshold=6, tilt=10, pm2_5=1e-5, pm10=1e-5)

~\Projects\pvlib-python\pvlib\losses.py in soiling_hsu(rainfall, cleaning_threshold, tilt, pm2_5, pm10, depo_veloc, rain_accum_period)
     67
     68     # accumulate rainfall into periods for comparison with threshold
---> 69     accum_rain = rainfall.rolling(rain_accum_period, closed='right').sum()
     70     # cleaning is True for intervals with rainfall greater than threshold
     71     cleaning_times = accum_rain.index[accum_rain >= cleaning_threshold]

~\projects\pvlib-python\venv\lib\site-packages\pandas\core\generic.py in rolling(self, window, min_periods, center, win_type, on, axis, closed)
  10145                                    min_periods=min_periods,
  10146                                    center=center, win_type=win_type,
> 10147                                    on=on, axis=axis, closed=closed)
  10148
  10149         cls.rolling = rolling

~\projects\pvlib-python\venv\lib\site-packages\pandas\core\window.py in rolling(obj, win_type, **kwds)
   2624         return Window(obj, win_type=win_type, **kwds)
   2625
-> 2626     return Rolling(obj, **kwds)
   2627
   2628

~\projects\pvlib-python\venv\lib\site-packages\pandas\core\window.py in __init__(self, obj, window, min_periods, center, win_type, axis, on, closed, **kwargs)
     63         self.win_freq = None
     64         self.axis = obj._get_axis_number(axis) if axis is not None else None
---> 65         self.validate()
     66
     67     @property

~\projects\pvlib-python\venv\lib\site-packages\pandas\core\window.py in validate(self)
   1576                                          timedelta))):
   1577
-> 1578             self._validate_monotonic()
   1579             freq = self._validate_freq()
   1580

~\projects\pvlib-python\venv\lib\site-packages\pandas\core\window.py in _validate_monotonic(self)
   1610             formatted = self.on or 'index'
   1611             raise ValueError("{0} must be "
-> 1612                              "monotonic".format(formatted))
   1613
   1614     def _validate_freq(self):

Expected behavior
I'm not sure, I guess I would expect it to be able to handle TMY3 data, but maybe it's the TMY3 with the year coerced to the same number that's the problem? This would also fail TMY3 if you didn't coerce the year, but it might fail silently in some weird ugly way that you didn't notice at first.

Versions:

• pvlib: 0.7.1+25.gc7b1c0e
• pandas: 0.24.2
• python: 3.6.8 (tags/v3.6.8:3c6b436a57, Dec 24 2018, 00:16:47) [MSC v.1916 64 bit (AMD64)]

Additional context
related to #865 and #866