Modify LILAC coupling to not need domain files

As mentioned here: https://github.com/ESCOMP/CTSM/pull/1258#issuecomment-802350060

> For now, lilac is still getting the mask/frac from fatmlndfrac, just like mct does. This will need some rework in the future if we want to avoid needing a domain file for the lilac case. (We talked before about just using the land mask / frac sent from the atmosphere at runtime, but now I'm not sure whether / how that would work, because we would need to pass this field in initialization, which could be tricky.)

As we noticed in a walkthrough of the state of the cap & xml files (with @mvertens and @ekluzek ), things are in a confusing state right now, because we're needing to keep xml variables for the LND_DOMAIN_FILE around in NUOPC just for the sake of testing LILAC. It would be good to fix this, but we need to figure out how the data flow should work for NUOPC. [Edit: should that say, "how the data flow should work for LILAC"?]