Hillslope hydrology may need changes to work correctly for virtual vegetated landunits

In discussions with @swensosc , we're not sure if the hillslope hydrology code will work correctly for virtual (zero-weight but active) vegetated landunits. As noted here https://github.com/ESCOMP/CTSM/pull/1715#discussion_r872752931

> The multiplication by col%wtgcell to create volumetric quantities highlights a potential issue here that we should at least discuss and possibly fix: I took pains a number of years ago to ensure that a zero-weight (virtual) vegetated landunit would operate effectively the same as a non-zero-weight landunit. This is needed for the glacier coupling and other aspects of dynamic landunits. Multiplying by col%wtgcell *might* be a problem in this respect. I think it depends on how these terms (stream_water_volume, qstreamflow, stream_water_depth and anything they affect) are used in the code. If they are just used for diagnostic purposes and to send gridcell average quantities to ROF, then I think this is fine. But if they feed back onto the column or landunit-level behavior of anything else in CTSM, then I think this is a problem: in this case, I'm thinking that a vegetated landunit that has zero weight on the gridcell would operate fundamentally differently because col%wtgcell would be exactly 0 for all of its columns, so these other terms would be exactly 0. So if these terms feed back onto other behavior, then I think we'll need to change the way these are specified so that they are on a per-area basis (like most other variables in the model).

This at least needs more thought, and may end up requiring some code changes to make this work right.