Add offset support to LogisticRegression for fixed/known linear predictor contributions

Hey, would it be possible to add support for an offset or fixed-coefficient parameter to LogisticRegression? This issue on Python's sklearn is exactly what I'm trying to do.

The API could look something like this:

// Pass a precomputed offset term per observation
// log_odds = offset[i] + β₀ + β₁·x[i]
let model = LogisticRegression::default()
    .alpha(1.0)
    .offset(offset_array)  // Array1<f64>, one value per training sample
    .fit(&dataset)?;

Start by locating the Rust implementation of LogisticRegression and its fit API, then compare the requested offset behavior with the linked scikit-learn issue. The work is done when LogisticRegression accepts one precomputed offset value per training sample and incorporates it into the stated log-odds model, with fitting behavior and API details defined.