hotfix: go for FeaturesType instead of InputData in a pipeline tuning

[Lopa10ko]

This is a 🙋 feature or enhancement.

Summary

• change the signature of the tune function so that the input_data param can be of type FeaturesType. This allows passing one of the following data formats: str, PathLike, numpy array, pandas DataFrame, FEDOT-native InputData, FEDOT-native MultiModalData, dictionary, or tuple
• perform several unit and integration tests to ensure the expected results
• add unit tests with different data formats (covered with test/unit/api/test_main_api and test/integration/pipelines/tuning/test_pipeline_tuning)

Context

related to discussion in #1298

[pep8speaks]

Hello @Lopa10ko! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2024-07-23 13:46:24 UTC

[github-actions]

All PEP8 errors has been fixed, thanks ❤️

Comment last updated at

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 3 lines in your changes missing coverage. Please review.

Project coverage is 79.94%. Comparing base (e0b4ee7) to head (7269af5).
Report is 1 commits behind head on master.

Files	Patch %	Lines
fedot/api/main.py	0.00%	3 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1311   +/-   ##
=======================================
  Coverage   79.93%   79.94%           
=======================================
  Files         146      146           
  Lines       10097    10102    +5     
=======================================
+ Hits         8071     8076    +5     
  Misses       2026     2026           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[aPovidlo]

Если подаются новые данные для тюнинга, то у них скорее всего будет новый и target. Поэтому в tune нужно по аналогии и target: TargetType = 'target' добавить

[Lopa10ko]

done

[aPovidlo]

@Lopa10ko Я мб еще учел если у нас input_data is None. то target должен быть также self.target. Или при вызове define_data он таким же и будет?

[Lopa10ko]

кажется, если input_data is None, то пост-тюнинг запускают на уже обработанных данных (на self.train_data), просто на этот раз дольше, например

а возможность запустить с указанными данными нужна в случае, если тюним на данных, которые не видели раньше (тогда и указываем FeaturesType + TargetType)