More accurate interpolation

[SouthEndMusic]

I have implemented the physical interpolation (as proposed in #225). I have not checked whether this implementation is exactly what we want for extrapolation (#279), but this implementation assumes that the storage is 0 at the lowest level and does linear extrapolation of the area as a function of the level based on the last 2 values.

This does not solve the problem of high oscillating flows produced by the ManningResistance node as hoped (#80). I think the problem there is that the calculated flows are way to large, and therefore the equilibrium is overshooted at every time step.

I cannot wrap my head around the current ManningResistance implementation. I understand that flow occurs due to a difference in level, but I do not understand why that means that all the water comes in motion, not just the bit at the top (which is why I find LinearResistance more intuitive). Is anyone familiar with a ManningResistance being used for flow that can go both directions, or is it only used for flow that goes one direction due to gravity? In that last case I understand better why all water is in motion.

Also, the current implementation assumes the presence of a lot of water in the channel itself, which fluctuates with the level in the upstream basin with no regard for (local) water preservation (by which I do not mean that the ManningResistance node is not preserving but it is less physical).

[SouthEndMusic]

I restricted this PR to the profile interpolation since fixing the ManningResistance turns out to be a separate problem.

[visr]

It looks good to me overall, I left some comments. Let's add some tests for get_area_and_level. For that I think you can create a Basin directly (not an entire test model), and then make sure you check all 3 main branches, including the area_diff 0 branches.