[BUG] DifferencingTransform inverse transform bug

[martins0n]

🐛 Bug Report

It seems DifferencingTransfrom doesn's handle df with multiple features.

For exampe in case below, I have error. But if i'd swapped LagTransform and DifferencingTransform -- the error would not have occurred

Expected behavior

It seems should work

How To Reproduce

from etna.datasets import generate_ar_df

tss = TSDataset(TSDataset.to_dataset(generate_ar_df(300, "2021-02-01", n_segments=20)), freq="1D")

linear_pipeline = Pipeline(
    model=LinearPerSegmentModel(), 
    transforms=[
        TimeSeriesImputerTransform(),
        LagTransform(in_column="target", lags=list(range(30, 100, 5))),
        HolidayTransform(),
        DifferencingTransform("target"),
        DateFlagsTransform(week_number_in_month=True, week_number_in_year=True, month_number_in_year=True),
        SegmentEncoderTransform()
    ], 
    horizon=24
)
metrics, forecast, info = linear_pipeline.backtest(tss, [MAE()], n_jobs=5)

Trace:

---------------------------------------------------------------------------
RemoteTraceback                           Traceback (most recent call last)
RemoteTraceback: 
"""
Traceback (most recent call last):
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/multiprocessing/pool.py", line 125, in worker
    result = (True, func(*args, **kwds))
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/joblib/_parallel_backends.py", line 595, in __call__
    return self.func(*args, **kwargs)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/joblib/parallel.py", line 262, in __call__
    return [func(*args, **kwargs)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/joblib/parallel.py", line 262, in <listcomp>
    return [func(*args, **kwargs)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/pipeline/pipeline.py", line 257, in _run_fold
    forecast = pipeline.forecast()
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/pipeline/pipeline.py", line 163, in forecast
    predictions = self.model.forecast(ts=future)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/models/base.py", line 25, in wrapper
    result = f(self, *args, **kwargs)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/models/base.py", line 141, in forecast
    ts.inverse_transform()
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/datasets/tsdataset.py", line 279, in inverse_transform
    self.df = transform.inverse_transform(self.df)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/transforms/math/differencing.py", line 369, in inverse_transform
    result_df = transform.inverse_transform(result_df)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/transforms/math/differencing.py", line 226, in inverse_transform
    result_df = self._reconstruct_test(df, columns_to_inverse)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/etna/transforms/math/differencing.py", line 181, in _reconstruct_test
    init_df.columns.set_levels([column], level="feature", inplace=True)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/pandas/util/_decorators.py", line 311, in wrapper
    return func(*args, **kwargs)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/pandas/core/indexes/multi.py", line 912, in set_levels
    idx._set_levels(
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/pandas/core/indexes/multi.py", line 795, in _set_levels
    new_codes = self._verify_integrity(levels=new_levels)
  File "/home/marti/.pyenv/versions/3.8.6/lib/python3.8/site-packages/pandas/core/indexes/multi.py", line 410, in _verify_integrity
    raise ValueError(
ValueError: On level 1, code max (15) >= length of level (1). NOTE: this index is in an inconsistent state

Environment

No response

Additional context

No response

Checklist

• Bug appears at the latest library version