Freamon

* Added date similarity support to flag_similar_records function:
* Implemented three date similarity calculation methods: linear, exponential, and threshold-based
* Added auto-detection of date columns using sophisticated heuristics
* Added handling for multiple date formats including native datetime, string representations, and timestamps
* Added configurable thresholds for date similarity calculations
* Created comprehensive tests for date auto-detection and similarity features
* Fixed duplicated parameter definitions in flag_duplicates.py

* Fixed AutoModelFlow cross-validation errors with similarity features:
* Modified Model.predict and Model.predict_proba to handle missing features by adding zeros
* Enhanced cross-validation to handle feature mismatches during training/prediction
* Added robust error handling around prediction and metrics calculation
* Improved handling of dynamic features like those from deduplication processes
* Added graceful fallback for metrics calculation failures

* Fixed additional issues with Model class compatibility:
* Fixed error in calculate_metrics() call with incorrect parameter order
* Added robust handling of metrics not explicitly requested but returned by calculate_metrics
* Added safety checks to prevent KeyError on log_loss metric
* Improved cross-validation metrics handling for different problem types
* Enhanced error recovery in model evaluation

* Fixed Model class compatibility with scikit-learn:
* Added get_params and set_params methods to comply with scikit-learn estimator interface
* Fixed TypeError in sklearn.base.clone when using cross-validation
* Added proper parameter handling for nested model objects
* Fixed integration between custom Model class and scikit-learn's cross-validation
* Improved error handling for parameter validation

* Fixed critical bugs in AutoModelFlow functionality:
* Enhanced topic model parameter filtering to explicitly define supported parameters
* Added better handling of custom parameters in topic modeling functions
* Fixed cross-validation implementation to properly call cv.split() method
* Added debug logging for filtered parameters to improve troubleshooting
* Added complete parameter validation for topic modeling
* Fixed pipeline integration when passing custom model parameters

* Added automatic text column detection to flag_similar_records function:
* Implemented smart detection of text vs. structured data columns
* Added specialized text similarity methods: fuzzy, TF-IDF, n-gram, and LSH
* Added text-specific thresholding and automatic weight boosting
* Fixed parallel processing issues for efficient handling of large datasets
* Added comprehensive test suite for auto-detection features
* Created detailed examples showing auto-detection capabilities
* Implemented backwards compatibility with existing code
* Added documentation for all new parameters and features