Fix water tracer values of downscaled precipitation

@swensosc 's hillslope hydrology (#1715) adds downscaling of precipitation terms. I think this will break the tracer values of these terms. We will need to fix this eventually if we want CTSM to be fully enabled for water tracers / isotopes and want that to work with hillslope hydrology.

First we should run a water tracer test (like SMS_D_Ld10.f10_f10_mg37.I2000Clm50BgcCrop.izumi_intel.clm-tracer_consistency) with hillslope on; I hope to see a failure in that test with hillslope on due to the need to fix these precipitation terms. (If we don't, then either I'm wrong about the need to fix this or the test isn't as powerful as I hope.)

To fix this, I think the best thing to do would be to have explicit fluxes of addition or subtraction of rain / snow from each column. (These fluxes could be calculated after the fact based on the difference between the column-level value and the gridcell-level value at the end of `downscale_hillslope_precipitation`.) Then calculate tracer versions of these fluxes and apply them to the tracer rain / snow amounts in each column before continuing with the current calculations of downscaled rain and snow in `SetDownscaledTracers`.

After making that change, we should verify that a tracer consistency test like the above one with hillslope enabled now passes.