Feature enhancements and bug fixes for generative text task

[kartik727]

A list of issues and potential enhancements for the newly added generative text task type

Bug Fixes

• Cache not being used in the compute_genai_metrics method of the RAITextInsights class.
• Cache does not appear to be updated from the front-end for the compute_question_answering_metrics and compute_genai_metrics methods of the RAITextInsights class.
• Error tree currently uses the model being evaluated to compute the AI assisted metrics instead of the evaluation metric.
• Additional text columns (like reference answers) in the input dataset not working as intended. As a result, metrics requiring reference answers (e.g., equivalence, exact match score, BLEU, etc.) are also currently not working.

Feature enhancements

• The templates with metric definitions for AI assisted metrics is currently hard-coded into the metric computation scripts. These can be made configurable in the future.
• Allowing the users to create new metrics by supplying their own templates with metrics definitions.
• Error tree - show name of the metric used for training the error tree (currently show mean squared error).
• Error tree - allow the use of other metrics for creating the error tree.
• Implement the front-end UI for model explainer.