autosens munges the carb sensitivity

[dburren]

Describe the bug
Our ISF describes how insulin affects our BG.
Our ICR describes how much insulin we need for carbs.
That's derived from old-pumping methods.
The more concise CSF is ISF/ICR (as used in the code already) and describes how each gram of carbs affects our BG.

When determine_basal adjusts the ISF based on sensitivityRatio (autosens) that makes sense as it's describing that insulin is affecting us more or less than "normal". However it does not also adjust the ICR, and consequently does adjust the CSF.

This seems fundamentally incorrect:
An increase in sensitivityRatio (needing more insulin) causes a decrease in ISF. Fine.
But with the current implementation it also results in a decrease in CSF, which is saying that carbs are somehow expected to affect our BG less by that same ratio.

Expected behavior
ICR to also be scaled by sensitivityRatio, so that csf is not affected.

oref0 0.7.1

[scottleibrand]

IIRC we tried that, and it worked less well, despite being theoretically more correct. I think it had something to do with how well carbs and insulin matched up when taken together at mealtime. I don't remember a lot of details, but I do remember we did a big simplification of autotune to more directly calculate CR (instead of trying to derive it as ISF/CSF) in May of 2017, if that helps for searching git commit history...