Shared by varpro_coverage() and gg_varpro(). out
needs z, tested and coverage; vars gives each
row's original variable, which the noise pattern and the kind are
read from. A row counts as tested here only when at least one tree scored
it as well as coverage >= trim, so trim = 0 cannot make an
unscored noise row a bar at 0. Adds noise, kind,
noise_cut and noise_ref (the kind of the noise that set the
bar, NA when none was tested), and sets selected.
.varpro_noise_judge(out, vars, object, noise, kind_values)out with the noise, kind, noise_cut,
noise_ref and selected columns set.