Add support for a transformers datasets load

[Lopa10ko]

Add transformation functions from DatasetDict to pytsbe-native long-formatted data source.

This issue is specifically aimed at avoiding the use of the "save to csv first and then upload files to pytsbe" workaround.

Example of `load_dataset` usecase

from datasets import load_dataset

configs = [
    'australian_electricity_demand',
    'bitcoin', 'car_parts', 'cif_2016',
    'covid_deaths', 'fred_md', 'hospital',
]

dataset_mapping = dict()

for config in configs:
    dataset_mapping[config] = load_dataset('monash_tsf', config)

from colab