Dynamic lakes will need to receive lake removed by ROF model

As a future development of dynamic lakes (see #1206) we'll need to receive the amount of water removed from lakes by the ROF model. Normally this will be just the reverse of the Precip minus Evaporation that's sent to ROF, but when lakes are empty or near empty it will be less than that. Since, it likely won't happen that often we figure we can ignore this effect initially. But, for completeness and in order to connect to an ocean model we'll likely need to put it in place.

Note, that there is a similar issue with glaciers that we ignore. And this is similar to how irrigation is currently handled in CTSM. We'll likely need to implement something similar to how irrigation, which means a field sent from ROF to CTSM to track it and then CTSM will need to react to the field sent back.

In mizuRoute this is documented as...

https://github.com/NCAR/mizuRoute/issues/170