Fixes dictionary registration of FilterEnsemble

[advieser]

closes #985
Dependent on mlr-org/mlr3filters#185

[advieser]

Can think on whether we want to pass some default values or NAs for the custom fields of our in-place R6 class, although looking at as.data.table(mlr_filters), this does not seem to be handled consistently at the moment in any case.

[mb706]

feature_types etc. should be "all" feature types, not none -- look these up from mlr_reflections

[advieser]

Looking entries up from mlr_reflections would lead to different entries depending on package load order.

library(mlr3proba)
devtools::load_all("./mlr3filters")
devtools::load_all("./mlr3pipelines")
as.data.table(flt())[key == "ensemble"]$task_types
#[[1]]
#[1] "classif"      "dens"         "regr"         "surv"         "unsupervised"

devtools::load_all("./mlr3filters")
devtools::load_all("./mlr3pipelines")
library(mlr3proba)
as.data.table(flt())[key == "ensemble"]$task_types
#[[1]]
#[1] "classif"      "regr"         "unsupervised"

[be-marc]

Sorry I overlooked this. You can close #990

[be-marc]

Fix for mlr3filters should be on CRAN soon

[be-marc]

New mlr3filters version is on CRAN