cross-validation: Malformed tables if shapefile with training points does not have consecutive labels

[Agustin Lobo]

Related to #300

(Version 3.6.20200407T152432.master) I confirm the classification and probability estimation in the classification workflow work fine, but there is a pending issue: the tables generated by the cross-validation include the non-existent classes.. eg: if classes are 1,2,3,11,12, the confusion table includes rows and columns for classes between 4 and 10.

‌

‌

[Andreas Janz]

Non-consecutive labels are now supported in the new Classification Workflow app.

Note that the Classification Workflow Classic app is not supporting it.

‌