snow fraction is uninitialized when passed to fates during cold-starts

### Brief summary of bug

On a cold-start with FATES, we are passing in an unitialized snow ground cover fraction:

Here is where the call happens:
https://github.com/ESCOMP/CTSM/blob/master/src/utils/clmfates_interfaceMod.F90#L1150-L1160

Then we call a routine that transfers these values into fates internal structures.
https://github.com/ESCOMP/CTSM/blob/master/src/utils/clmfates_interfaceMod.F90#L1164

On the cheyenne intel compiler, the snow fraction was 1e+35. The snow depth seemed to have a normal sensible value. In fates we multiply the snow cover fraction by the mean snow depth to get an effective column snow depth. My suspicion is that for some sites in upper latitudes, we end up simulating a time-step with everything under an unfathomable amount of snow, which melts on the next time-step. No big deal?

I found this issue while working on the two-stream code, having no exposed vegetation triggered an error there.