There are two steps: (a) use the package manager to load all Java code from all WEKA packages that are installed, including the RPlugin (which obviously needs to be installed first using the package manager), and (b) use reflection to create a classifier object.
Here is an example using xgboost to build the classifier via MLR. You can run this code in WEKA's Groovy console (available via the Tools menu in the GUIChooser), but this should also be valid Java once wrapped into a class.
If xgboost has not yet been installed in R as an R library, there may be some delay while the library gets installed through R in the background.
// Load all packages so that MLRClassifier can be found