Evaluate temperature disaggregation in polar regions

[arbennett]

While reviewing #208 I made some plots to see how the temperature disaggregation looked, and think maybe some further heuristics could create more physically realistic temperature timeseries. For this test I just used some temperature data for another location and set the latitude to 75. This explains the larger than realistic diurnal temperature ranges, but was just used for exploration.

From the figure we can see that during both periods a very sharp temperature gradient is set either mid-day or end-of-day, which may not be realistic. Two other options seem like reasonable alternatives:

1. Set the temperature disaggregation so that during these periods is as smooth as possible, but with the relative times of min/max temperature flipped between all daylight and no daylight periods
2. Use the solar angle as a guide for when min/max temperatures should occur. This may not make sense in the case of no-daylight since the solar angle will always be negative, but keeps the transition consistent between time periods.

[arbennett]

Setting rise times and set times to 4AM and 4PM produces this as an alternative: