Feature request: method to save fitted models

[sammiller06]

Description

StatsForecast currently lacks a method to save fitted models to disk.

Currently pickling the models that I access with the following snippet:

fitted_models = sf.fit().fitted_

Use case

This would be helpful in use cases where we want to call predict() separately on a pre-trained model.

[philippschmalen]

You could use joblib to store model artifacts.
Here is an example

import joblib
from statsforecast import StatsForecast
from statsforecast.models import MSTL, SeasonalNaive

season_length = 24

sf = StatsForecast(
    df=df,  # assume you have a dataframe loaded
    freq="H",
    n_jobs=-1,
    models=[
        MSTL(
            season_length=[
                season_length,
                season_length * 90,
            ],
        ),
    ],
    fallback_model=SeasonalNaive(season_length=season_length),
)

sf.fit()

# store StatsForecast predictor in joblib
model_filename = "model/mstl.joblib"
joblib.dump(sf, model_filename)

# load from file
model = joblib.load(model_filename)
forecast = model.predict(h=30, level=[90])

[jmoralez]

I believe this was fixed by #667