Feature/multilingual textstats

[jmsevin]

This pull request introduces several improvements, including enhanced language support for readability and duration prediction, updates to the setup process in the documentation, and refinements to default behavior for reading speed calculations. Below is a breakdown of the most significant changes:

Language Support Enhancements:

• welearn_datastack/constants.py: Added reading speeds and Flesch-Kincaid constants for new languages (de, es, jp, pt, ar, it, nl, zh) to improve language-specific readability and duration predictions. [1] [2]
• welearn_datastack/utils_/text_stat_utils.py: Updated the lang parameter descriptions in predict_readability and predict_duration functions to reflect the newly supported languages. [1] [2]

Documentation Updates:

• README.md: Updated the setup instructions to use Poetry for dependency management instead of a virtual environment with pip. This simplifies the setup process and aligns with modern Python project practices.

Default Behavior Refinements:

• welearn_datastack/utils_/text_stat_utils.py: Changed the default reading speed in predict_duration to the average of all supported reading speeds (184 wpm) instead of defaulting to English. This ensures a more balanced fallback for unsupported languages.