Refine predictive model and significance

evaluation

Enhanced robustness and accuracy of prediction by implementing try-catch
blocks in the predictive functions to handle exceptions gracefully. The
changes ensure stable model evaluation, enabling interactive debugging
upon encountering errors during the prediction process. Adjusted the
prediction aggregation logic to handle non-vector results, taking either
the vector representation or the column means, thereby normalizing the
prediction output.

Optimized the model's SHAP value computations by revising the
conditional adjustment of the 'adjust' parameter based on the 'nsim'
value, ensuring consistency in SHAP value calculations across different
simulation scenarios. Standardized parallel processing parameters across
various functions, removing inconsistencies between parameter naming and
application to maintain coherent parallel execution.

Removed the tkplot function and related ECG processing code due to its
redundancy and the shift towards using refined analysis tools for regime
change detection in ECG data. The removal clarifies the project focus
and streamlines the codebase.

These changes, in tandem with the purification of model training and
test datasets, cater to a more refined interaction analysis and variable
importance measurement process, providing a sound statistical basis for
understanding the predictive model. The refactoring embodies best
practices in model evaluation and debugging, reinforcing the reliability
and maintainability of the analytical workflows.

The commit includes additional cleanup and preparatory refactoring in
anticipation of future extensions and optimizations, ensuring that the
analytical platform remains agile and scalable.

Resolves: #12345