Add local.importance parameter in ranger learners

[gi-marengo]

The ranger package supports the parameter local.importance = TRUE to compute casewise variable importance values.
This is very useful for model interpretability, but currently the parameter is not available in the ranger learners.
Is possible to add it?

I also tried to add the parameter manually as suggested in the mlr3 learner class documentation:

# Example provided in the mlr3 learner class documentation for classif.rpart

lrn = lrn("classif.rpart")
lrn$param_set$add(paradox::ParamFct$new("foo", levels = c("a", "b"))) 

lrn$param_set$add(paradox::ps(foo = p_fct(c("a", "b")))) # also with the syntax of paradox 1.0.1

However, in the current paradox version the $add() method is no longer available, leading to

> Error: attempt to apply a non-function.

and I didn’t find an alternative way to add a new parameter to an existing learner.