Design protocol classes and update architecture for model comparison analysis

[EH-MLS]

To expand our analysis capabilities, we want to introduce model comparison analysis in addition to existing segmentation and drift analysis.

Requirements:

• Support comparison of multiple sets of prediction data (no direct model interaction, only prediction data).
• Introduce two protocol classes:
  • Prediction Calculation Protocol:
    • Should be initialized with a DataFrame and a label column name.
    • The __call__ method should accept two arguments:
      1. a list of prediction Series (pred-data)
      2. an optional function for calculating pairwise error (defaulting to a standard metric if not provided)
    • The protocol should calculate pointwise error for each pred-data against the label, using the specified function.
  • Plotting Protocol:
    • Responsible for visualizing the comparison results between multiple predictions and the label.
• Update the overall architecture to integrate these protocol classes and allow extensibility for new analysis types.

Expected Outcome:

• Specification of the two protocol classes, including their responsibilities and interface definitions.
• Updated architecture design or diagram showing how model comparison integrates with the existing pipeline and interacts with prediction data.
• Guidance for supporting multiple predictions and extensibility for future analysis modules.