Isotonic Regression

Each algorithm is implemented as both a function that mutates a vector of regressors with an optional weight vector or as a non-mutating version of the same function. As is idiomatic in Julia, we denote the mutating versions by a an exclamation mark (!).

isotonic_regression(y::Vector{Float64}, weights::Vector{Float64})
isotonic_regression!(y::Vector{Float64}, weights::Vector{Float64})

There are additional overloads for the case where the weight vector is simply the ones vector.