introduce and use weighted_maybe/2

[triska]

Many thanks to @hurufu for the very nice definition!

This is until library(random) provides this functionality, and a good name is found for it. See #2670.

[hurufu]

Informal proof of correct probabilities:

% Given that numbers from interval [0,N) are equally likely, X has N possible
% values with 1/N probability. Observe that for every non-negative value K ≤ N:
% if K = 0 then P(X < 0) = 0   (No value X exist)
% if K = 1 then P(X < 1) = 1/N (when X = 0)
% if K = 2 then P(X < 2) = 2/N (when X = 0 or X = 1)
% ...
% if K = N then P(X < N) = 1   (for all possible values of X)
%
% Which exactly corresponds to our expectation, that goals succeeds with
% probability K/N.