Add dataframe format converter to plot_holidays

[alex-hse-repository]

🚀 Feature Request

Add dataframe format converter to plot_holidays

Motivation

Make method more convenient for parameters selection

Proposal

1. Create method prophet_to_etna(holidays: pd.Dataframe) -> pd.Dataframe
   Convert the format of holidays dataframe from Prophet see to the format of plot_holidays
2. Create method etna_to_prophet(holidays: pd.Dataframe) -> pd.Dataframe
   Do it vice versa

Test cases

No response

Alternatives

No response

Additional context

No response

Checklist

• I discussed this issue with ETNA Team

[iKintosh]

Why not just make this format default to the library?
We have this exact format for ProphetModel already. Why should we support two different holiday formats?

[iKintosh]

Actually prophet_to_etna(holidays: pd.Dataframe) -> pd.Dataframe is not possible, since we have to know exactly start and end of TSDataset with data we want to apply these holidays to. Additionally we have to know frequency of TSDataset to build correct holiday_df.

[iKintosh]

I think it should be done like this:

import pandas as pd
from etna.datasets import TSDataset
from etna.analysis import plot_holidays

ts = TSDataset(data_in_etna_format, freq="D")
new_years = pd.DataFrame({
  'holiday': 'New Year',
  'ds': pd.to_datetime(['2015-01-01', '2016-01-01', '2017-01-01', '2018-01-01', '2019-01-01']),
  'lower_window': 10,
  'upper_window': 10,
})
holidays = pd.concat((new_years,))

plot_holidays(ts, holidays=holidays, segments=["Finland_KaggleMart_Kaggle Hat"], columns_num=1, figsize=(10,7))

[iKintosh]

However I suggest we add converter function from old etna format to current (prophet) format.