Constrain z scale in some meaningful way

[xmacex]

Especially precipitation data can range quite dramatically, whereas the abstract function as it is now is something like 𝛘² with k = 3, a skewed gaussian or something. Anyway the precipitation is zero or close to zero for many days, but e.g. 2017-09-06, exactly one month before Lucy Suchman was at IT University of Copenhagen, a daily precipitation value was > 20 mm.

So this might happen, as seen in preview of OpenSCAD. The left one is unconstrained with the above data and an arbitrary z scaling, middle one uses a cropping function, and the one on the right uses a local z maximization (local in the space of all potential data objects) so that maximum value is brought to the available range, and then other points are scaled linearly. Maximizing z range hinders comparability across objects.

One idea is simply to exploit smoothing to bring these sorts of values down. Here, with alpha 25 the extreme peaks are almost under control, there is still a little bit of clipping.

Here's midsummer 2015, with > 30mm daily precipitation.

I think we agreed to have 4 precipitation measurements per day, available from some data source... was it AccuWeather? That shouldn't cause issues, but there had better be a disciplined approach anyway.

[xmacex]

On 2018-08-28 we agreed with Baki to go for clipping.

[xmacex]

Commits with clipping https://github.com/xmacex/make_a_data_object/commit/2582af70744394d20cfe83ef9f6db5281db98df8 and introduction of additive surface function https://github.com/xmacex/make_a_data_object/commit/4bcde1b2c6eaa24f90f824b6a291abe1f9daf7a4 conjunctively solve this sufficiently well.