Should decomposition functions reuse GHI index for DOY if none?

[mikofski]

As discussed in #1179 many decomposition functions like Erbs require both GHI as some numeric type and a separate DOY or datetime series datetime_or_doy which can be either integer, float, numpy ndarray or pandas DatetimeIndex. It's likely that many users will pass GHI as a pandas series, and so it's possibly an extra burden to also pass datetime_or_doy, which could be simplified by reusing the index of GHI. EG:

if datetime_or_doy is None:
    try:
        datetime_or_doy = ghi.index
    except AttributeError:
        raise TypeError('datetime_or_doy can't be None if ghi has no index')

If there's consensus we can update all decomposition methods in a single PR.