Fix actual return type of column_names property

[glatterf42]

Uncovered another bug introduced by #156 and #157: previously, I returned the column_names property as a cast(list[str], ...) of the underlying sqlalchemy collection. Mypy understood this just fine, but the actual type of the return object was not changed, it was just made to look that way. This created an error in pd.DataFrame().set_index() as this function does not know how to handle sqlalchemy's _AssociationList objects.
This PR fixes this bug by explicitly creating the return object as a list[str]. The for loop should be fine as no item should have more than 12 column_names in practice. It also adds tests to ensure this behaviour doesn't break again in the future.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.5%. Comparing base (ef26704) to head (7ca6c9d).

Additional details and impacted files

@@                        Coverage Diff                         @@
##           fix/column-names-for-same-indexset    #162   +/-   ##
==================================================================
  Coverage                                88.5%   88.5%           
==================================================================
  Files                                     223     223           
  Lines                                    8113    8117    +4     
==================================================================
+ Hits                                     7185    7189    +4     
  Misses                                    928     928           

Files with missing lines	Coverage Δ
ixmp4/data/db/optimization/equation/model.py	100.0% <100.0%> (ø)
ixmp4/data/db/optimization/parameter/model.py	97.4% <100.0%> (+<0.1%)	⬆️
ixmp4/data/db/optimization/table/model.py	100.0% <100.0%> (ø)
ixmp4/data/db/optimization/variable/model.py	100.0% <100.0%> (ø)