pvfactors_timeseries docstring types are incorrect

[kandersolar]

Describe the bug
The pvfactors_timeseries docstring currently lists several parameters as "numeric", implying scalar values are acceptable (docs). However, some parameters cannot actually be scalars and should be listed as "array-like" instead.

To Reproduce
Example showing that scalar inputs for the module orientation parameters are not acceptable:

Click to expand!

import pvlib
import pandas as pd
from pvlib.bifacial import pvfactors_timeseries

times = pd.date_range('2019-01-01', '2019-01-02', freq='15min').tz_localize('Etc/GMT+5')
location = pvlib.location.Location(40, -80)
solarpos = location.get_solarposition(times)
clearsky = location.get_clearsky(times)

kwargs = dict(
    solar_azimuth=solarpos['azimuth'],
    solar_zenith=solarpos['apparent_zenith'],
    axis_azimuth=0,
    timestamps=times,
    dni=clearsky['dni'],
    dhi=clearsky['dhi'],
    gcr=0.3,
    pvrow_height=1.5,
    pvrow_width=1.0,
    albedo=0.2
)

pvfactors_timeseries(surface_tilt=20, surface_azimuth=90, **kwargs)  # IndexError: too many indices for array
pvfactors_timeseries(surface_tilt=pd.Series(20, index=times),
                     surface_azimuth=pd.Series(90, index=times),
                     **kwargs)  # works

Expected behavior
Correct the types in the docstring to accurately reflect pvfactors' requirements.

Additional context
Originally discussed here: #1331 (reply in thread)
Edit: this is more or less a duplicate of #1127

[campanelli-sunpower]

@kanderso-nrel My understanding is that array-like should accommodate scalars, in which case I think this means pvfactors is advertising falsely. Thoughts?

https://numpy.org/devdocs/reference/typing.html#numpy.typing.ArrayLike

[kandersolar]

Oh right, I had forgotten about the inconsistency between how pvlib and numpy each define array-like. #765 is very relevant. Maybe we call it array-like for now until we change the pvlib convention later? I'd also be fine with array or Series, or something like that.

[kandersolar]

Maybe this is less a docs bug and more a functionality bug. I generally expect pvlib functions to work with scalar orientation values. A better fix may be to allow scalar inputs and to cast scalars to arrays inside the function prior to forwarding them on to pvfactors.

[cwhanse]

A better fix may be to allow scalar inputs and to cast scalars to arrays inside the function prior to forwarding them on to pvfactors.

+1