wrapper: wrapper greedy search for structure
In gherardovarando/Rbnet: Define and Manage Bayesian Networks

wrapper greedy search based on evaluation of training-error for general funciotn that define a structure

1 2	wrapper(targets, predictors, data, modelfun = naive_bayes.bnet, bnet = NULL, evalfun = "MSE", exit = F)

`targets`	vector of names for target variables
`predictors`	vector of names for target variables
`data`	data.frame of observations
`modelfun`	function, function that define bnet must accept data, targets and predictors as input and output a bnet
`bnet`	bnet object, starting strucure or NULL
`evalfun`	string, one of "MSE" or "MAE", or other string naming a scoring function
`exit`	logica, if `TRUE` teh search stops at first not improvement of the score