Better control on the inputs of robustness_categories

[RondeauG]

Addressing a Problem?

Currently (as of #299), the output of robustness_fractions is given directly to robustness_categories. This means that the function will use changed and agree to output the agreement on the direction (strong/no change/conflicting), but not the actual direction itself.

Potential Solution

Instead of agree, I think that sending agree_positive would accomplish what I need. This could be an argument somewhere, either here in xscen or in xclim.

Additional context

I need to be able to compute the "Direction" categories used in the Hydroclimatic Atlas, with a potential 6th category for "Conflicting signals": https://cehq.gouv.qc.ca/atlas-hydroclimatique/carte-indicateurs/index.htm

Contribution

• I would be willing/able to open a Pull Request to contribute this feature.