Consider using the IOB curves from oref0/openaps

[trixing]

LoopKit uses a static model for only 4 supported DIAs.

I'd propose replacing the algorithm with the curve found in oref0/openaps:

https://github.com/nightscout/cgm-remote-monitor/blob/master/lib/plugins/iob.js
https://github.com/openaps/oref0/blob/master/lib/iob/calculate.js

This gives a much smoother curve at the end of the action duration.

Especially visible in the derivatives of the functions

[ps2]

The discontinuity in the end of the graph does not have much effect on treatment decisions. However, we are looking at the timing of the peak, based on graphed plots of insulin counteraction effects in LoopKit/Loop#507.

@scottleibrand I seem to remember you mentioning that OpenAPS has a peak at around 75 minutes., but the graph above shoes a peak at 100 minutes. Is the graph above wrong, or perhaps it was for a different DIA?

[scottleibrand]

The 75 minute peak is with a DIA of 3 hours. With 4 hours it's stretched to 100 minutes.

[ps2]

Not going to use the curves from OpenAPS. Work on using separate exponential curve model here: LoopKit/Loop#539