Make qg (surface humidity) calculation consistent across landunits

Currently, qg - surface humidity - is calculated differently for vegetated vs. special landunits. For vegetated landunits, separate qg values are computed for the soil, snow and h2osfc-covered portions of each column; for special landunits, a single qg value is computed based on the weighted average temperature of these different fractions.

I'd like to make the special landunit calculation of qg follow the method used for vegetated landunits. This is partly for the sake of consistency in the physics, and partly because I think this will make it easier to implement the tracer form of qg.