[BUG] SARIMAX can't predict interval if horizon=1

[Mr-Geekman]

🐛 Bug Report

SARIMAX can't predict interval if horizon=1, it fails with an error:

AttributeError: 'int' object has no attribute 'startswith'

The problem is inside _SARIMAXAdapter.predict. If horizon=1, then returned forecast after casting to pd.DataFrame hasn't "predicted_mean" column.

Expected behavior

No error.

How To Reproduce

import pandas as pd

from etna.datasets import TSDataset
from etna.models import SARIMAXModel
from etna.pipeline import Pipeline

HORIZON = 7


def main():
    df = pd.read_csv("examples/data/example_dataset.csv", parse_dates=["timestamp"])
    ts = TSDataset(df=TSDataset.to_dataset(df), freq="D")
    train_ts, test_ts = ts.train_test_split(test_size=HORIZON)

    # sarimax
    sarimax_pipeline = Pipeline(model=SARIMAXModel(), quantiles=[0.025, 0.1, 0.4, 0.9, 0.975], horizon=1)
    sarimax_pipeline.fit(ts=train_ts)
    forecast_sarimax = sarimax_pipeline.forecast(prediction_interval=True)

if __name__ == "__main__":
    main()

Environment

No response

Additional context

No response

Checklist

• Bug appears at the latest library version