"presolved" solved status is misleading for instances that were solved during feature computation

[larskotthoff]

There are basically two ways instances can be solved before the actual selection starts:

• A presolver is run for a set amount of time.
• The computation of a feature value causes the problem to be solved, e.g. probing features or unit/root node propagation.

Both of these cases are currently treated in the same way -- the value in feature_runstatus would be "presolved". I propose have different values, i.e. "presolved" if an actual presolver was run, else "featuresolved".

This difference is relevant because a particular model may not use the feature whose computation caused the problem to be solved. In this case, an instance would remain "presolved", but no longer "featuresolved".