[BUG] nan in target for forecasting with prediction intervals for SARIMAX

[martins0n]

🐛 Bug Report

Nan in targets of SARIMAX predictions with prediction intervals enabled. We have predicted_mean column instead.

Expected behavior

We expected forecasts in target column, not predicted_mean

How To Reproduce

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from etna.analysis.plotters import plot_forecast
from etna.datasets import TSDataset

url = r"https://raw.githubusercontent.com/owid/covid-19-data/master/public/data/owid-covid-data.csv"
df = pd.read_csv(url)
usecols = ["date", "location", "new_cases_per_million"]
countries = ["Russia", "United States", "Germany", "Canada", "Italy", "Australia"]
df = df[df.location.isin(countries)][usecols]
df = df.sort_values(["date", "location"]).reset_index(drop=True)
df = df.rename(columns={"date":"timestamp", "location":"segment", "new_cases_per_million":"target"})
df = TSDataset.to_dataset(df)
ts = TSDataset(df, freq="D")

pipeline = Pipeline(model=SARIMAXModel(), transforms=[], horizon=30)
pipeline.fit(ts)
forecast_df = pipeline.forecast(prediction_interval=True)
forecast_df.head()

Environment

No response

Additional context

No response

Checklist

• Bug appears at the latest library version