lambda regularisation parameter not possible to customise for variable selection of outcome models

[LukaszChrostowski]

Currently, the package supports variable selection under selection (ipw) models implemented using the Rcpp package and under outcome (mi) models using the cv.ncvreg function of the ncvreg package. Both methods should allow the user to run the model with a customised version of lambda (regularisation parameter) - the default is selection with cross-validation algorithm.

Since control_selection list allows to define it for selection model (lambda argument), control_outcome does not (lambda argument is missing), so CV algorithm is executed each time, I suggest to add this possibility by running ncvreg function instead of cv.ncvreg when lambda in control_outcome list is set. This should be implemented in nonprobMi.R file (lines 190-207)