Soil problems

This issue is a follow up of issue #874 (closed). Further work with soil and relatively high ice/water/voids fractions have shown more issues. More synthetic simulations have been designed in order to identify where the problems might come from.

• One such simulation consists in setting all parameters constant (TA=0°C, ISWR=0, ILWR=OLWR, PSUM=0) with a soil made of a single layer set at 0°C and with many elements. When running with bucket, besides a slow freezing coming from the soil surface and diffusing into the soil, everything is constant, as expected. When running with Richards, there are temperature oscillations as well as soil water content variations. See the required files for the simulation (it relies on the new MeteoIO's plugin "SYNTH") io_all_cst.ini, stb2.sno as well as a result when running with Richards:
• Another simulation builds on the first one but generates a sudden warming and precipitation event (the warming is there to allow for liquid precipitation). In this case, we see the same artifacts during the constant phase and then a behavior totally different from what was simulated with the old-trunk branch of Snowpack (ie before the merge of the water vapor transport modeling). See the files required for the simulation: io_sudden_precip.ini, stb2_warm.sno and a comparison plot: