Elegantly handle radiance

`vismodel()` et al. currently need to be 'tricked' into working with radiance data. In principle it should be fine to use radiance measures directly as backgrounds or even stimuli, rather than modelling them (illum x reflectance), but will just need some thought.