Add support for h2o package

[bgreenwell]

Some support for regression models is there, but it needs to be tested! Also, h2o.varimp() output is model specific!

[bgreenwell]

Here is what h2o is doing internally (will need to check for NULL):

#'
#' Retrieve the variable importance.
#'
#' @param object An \linkS4class{H2OModel} object.
#' @export
h2o.varimp <- function(object) {
  o <- object
  if( is(o, "H2OModel") ) {
    vi <- o@model$variable_importances
    if( is.null(vi) ) { vi <- object@model$standardized_coefficient_magnitudes }  # no true variable importances, maybe glm coeffs? (return standardized table...)
    if( is.null(vi) ) {
      warning("This model doesn't have variable importances", call. = FALSE)
      return(invisible(NULL))
    }
    vi
  } else {
    warning( paste0("No variable importances for ", class(o)) )
    return(NULL)
  }
}