Add Flexible Missing Data Imputation Functionality

[shubhamnaikk]

Description

This pull request introduces a robust and flexible impute_missing_values function to the Koheesio framework, enabling users to handle missing data efficiently. The function supports multiple imputation strategies, providing versatility for different use cases.

##Key Features:
Imputation Strategies:
Mean: Replaces missing values with the column's mean.
Median: Replaces missing values with the column's median.
Mode: Replaces missing values with the column's mode.
Constant: Allows users to replace missing values with a specified constant.
Error Handling:
Raises a ValueError if a specified column does not exist in the DataFrame.
Validates the imputation strategy and ensures the fill_value is provided for the "constant" strategy.
Scalable Design:
Designed to handle multiple columns at once, making it efficient for large datasets.
Flexible to extend with additional strategies in the future.

Tests Included:

Comprehensive unit tests have been added to validate the functionality and ensure robust error handling:

Mean Imputation: Validates the replacement of missing values with the mean.
Median Imputation: Checks correctness for median-based replacement.
Mode Imputation: Ensures mode-based imputation works as expected.
Constant Imputation: Tests custom value replacement for missing values.
Error Handling: Verifies that invalid columns raise appropriate errors.

Impact:

This feature significantly enhances the data preprocessing capabilities of the Koheesio framework.
It provides a flexible and reliable method to handle missing data, a critical step in building robust data pipelines.

[dannymeijer]

Thank you for your contribution @shubhamnaikk , really appreciate it.

A few points:

• The placement of your modules does not quite fit with how we intend, it would be better if we placed these elsewhere;
• The way you wrote the code, it is not making use of the Koheesio framework (Step classes) along with the typing that we intend to enforce.
• Currently we have no module for pure python implementations (Transformation in your case) - we should add that as part of your PR if that is the intended use.

Also, can you explain the intended use for this a bit more? Would this be for ML use-cases with the input being something like pandas perhaps? Or did you have something else in mind.

I propose that we have a small meetup to discuss, as I would love to add your contribution to our library. Please reach out to me in a DM / email so we can discuss further.

[dannymeijer]

Please also see: #129

[dannymeijer]

There has not been any response for the last 2 weeks. Please respond or address the aforementioned concerns before the end of the week.