projection baseline offset in warming levels determination

[perrette]

This issue is not really an issue, but here for documentation. So you can object if needed.

As I am reviewing existing code, I see the GMT in ISIMIP GCMs is adjusted for observed temperature, i.e. GMT - GMT[baseline].mean() + observed_baseline, e.g. 0.86 for 1995-2014 or 0.63 for 1986-2005. This means we actually use "Warming above baseline" rather than "Warming above pre-industrial". This has the advantage that when running the emulator with well-calibrated temperature, a value in 2030 will resemble the input GCM values in 2030. I added the following description in theget_matching_years_by_time_bucket docstring.

Next to the warming levels and year, the output contains the actual warming, rate of warming and accumulated warming.
Note that while the accumulated warming is calculated on the temperature above pre-industrial,
it is zeroed at the projection baseline, mostly for practical purpose, because we only downloaded the data since 1980s.
(similarly the GMT above PI is also aligned to the projection baseline, using observed warming during the baseline period)
The advantage of this approach is that the models align with each other for the present-days. The trade-off is that we
"erase" the memory since 1850, and any time-lagged effect of the warming inside these models.

I just wanted to make that clear @NiklasSchwind, @cschleussner. I don't have a strong feeling for this. In case you do, please comment here.