Ignore the modelled irradiance, use pvlib

[numerobis]

Most of the irradiance values (ghi/dni/dhi) are modelled using a model from 1990, and in some locations they're missing at various times of year (e.g. Montreal is missing all but the summer; Baker Lake is missing the winter). I haven't checked; they may be missing at other times too.

pvlib has the ability to model them, so we should:

1. check if the value is modelled ('M' in the notes field)
2. if so, parse all the weather data and feed it to the pvlib forecast models