Symbolic regression variable impact calculation does not work for cross-validation runs

[gkronber]

r6709:

• symbolic regression variable impacts can be calculated for each fold separately for cross-validation runs.
• Added a drop down box in the variable impact view to choose the fold.
• Minor change: variable impacts are rounded to 3 decimal digits.

[gkronber]

r6766: added possibility to calculate impacts over all folds for cross validation runs.

[gkronber]

r6775: fixed a crash when trying to calculate variable impacts for runs that did not collect statistics about variable frequency.

[gkronber]

r6783: fixed a bug leading to incorrectly sorted row names in the variable impact view.

[mkommend]

Comments:

• An exception occurs if a run does not contain results for all folds (e.g.,it has been stopped before the execution was finished).
• If a RunCollection contains CV runs with a different number of folds, no variable impacts can be shown.
• The fold combo box should be invisible, when no CV runs are analyzed.
• If CV runs are mixed with 'normal' runs in a RunCollection, the results for the 'normal' runs are not shown.
• The labeling of the runs changes when the value of the folds combo box gets changed.

[gkronber]

r7179: fixed the problems highlighted by the review of mkommend.

• cleaned up code
• added a check if there is a mix of normal and CV runs
• added a check if all runs contain a variable impact result
• added a check if all runs contain variable impacts for the same input variables
• added code to hide and show the folds combobox and its label
• changed the column names to show the same names when all folds are selected and when only specific folds are selected.

[mkommend]

Thx for correcting these issues.