fix ar order selection

[mathause]

• Closes #xxx
• Tests added
• Passes isort . && black . && flake8
• Fully documented, including CHANGELOG.rst

I was playing around with #109 when I noticed some tests randomly failing. Took me too long what's going wrong. Writing a test for this would be a pain...

One thing I noticed - the way it is set up, we select AROrder = 1 for the following situation:

                          order
scenario ensemble_member       
hist     r1i1p1f1             1
         r2i1p1f1             2
ssp126   r1i1p1f1             2
         r2i1p1f1             2

arr.quantile(q=0.5, method="nearest") over the scenarios selects [1, 2] and from this we select 1. (However, rounding up is no better, because then we could end up selecting 2 for (3 x 1 and 1 x 2). I may have to open an issue for this. Maybe it's also only relevant in this artificial scenario...

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (e75aa35) 87.00% compared to head (9a86a8a) 87.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #305   +/-   ##
=======================================
  Coverage   87.00%   87.00%           
=======================================
  Files          38       38           
  Lines        1601     1601           
=======================================
  Hits         1393     1393           
  Misses        208      208           

Flag	Coverage Δ
unittests	87.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
mesmer/calibrate_mesmer/train_gv.py	94.91% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[mathause]

I got this wrong in #176 which is not released.