Improve the user interaction.

We have simplified the user interaction, but also allow the user to continue with the classification even if the init.samples and number of models was not sufficient. With this version the user has the change to either manually increase it, automatically increase it, or abort the classification.

The legend of the output leaflet now shows percentage range and in the title has the number of models (nb_models value) which succeeded the sampling phase.

For many functions we have set the package.

We have also set a clause that checks if the results directory is empty. This avoid the error when plotting results because the threshold is defined based on the number of .tif images in the results directory.