Cross-Validated glmnet Procedures as Default Learners

[PhilBoileau]

unicate() and sunicate() rely on default sl3 super learner objects to estimate nuisance parameters. While there are compelling statistical reasons for estimating these parameters in this way, it would be more computationally efficient and less burdensome on users not familiar with the super learning methodology to use cross-validated glmnet routines as default estimators instead. This would also allow uniCATE to be submitted to CRAN. This isn't currently possible because sl3, an R package exclusively hosted on Github, is imported by uniCATE. Note that this change would not affect the asymptotic properties of methods implemented in unicate() and sunicate().