Bugfix for incorrect dimension-accessors when initializing from pandas

[danielhuppmann]

Please confirm that this PR has done the following:

• Tests Added
• Documentation Added
• Name of contributors Added to AUTHORS.rst
• Description in RELEASE_NOTES.md Added

Description of PR

This PR introduces an explicit removal of unused levels when initializing from a pandas object. This bug could lead to inconsistent return-values of the dimension-accessors.

closes #762

[codecov]

Codecov Report

Merging #763 (98f59f0) into main (4bfcec0) will increase coverage by 0.0%.
The diff coverage is 100.0%.

@@          Coverage Diff          @@
##            main    #763   +/-   ##
=====================================
  Coverage   94.4%   94.4%           
=====================================
  Files         59      59           
  Lines       6080    6086    +6     
=====================================
+ Hits        5744    5750    +6     
  Misses       336     336           

Files Changed	Coverage Δ
pyam/utils.py	92.7% <100.0%> (+<0.1%)	⬆️
tests/test_core.py	100.0% <100.0%> (ø)

[znicholls]

lgtm, one suggestion for a more defensive way to write the test (not an essential change)

[phackstock]

Beautiful thanks a lot @danielhuppmann 🎉

[coroa]

ok, thanks, indeed, that is the issue i encountered today.

i would have suggested to replace calls to get_index_levels method with its pandas version, ie something like:

def get_index_levels(index, level):
     return index.unique(level)

so they can decide what is the most efficient solution, but your fix is also fine.