Add errors section, fix dynamic evals date issue

[amrit110]

This pull request introduces a monthly error analysis feature to the evaluation dashboard and adds supporting backend and utility scripts for improved forecasting evaluation. The main changes include a new backend API endpoint for monthly metrics, frontend integration with visualization and error highlighting, and utility scripts to support unbiased evaluation data management.

Monthly Error Analysis Feature

• Added a new backend API endpoint /evaluation/static/monthly to provide monthly RMSE/MAE error metrics, enabling the identification of seasonal patterns and months with highest/lowest errors.
• Integrated monthly analysis into the frontend: fetches data from the new endpoint, displays a summary of best/worst months, trend analysis, and a line chart of monthly errors in EvaluationPage. [1] [2] [3]
• Defined new TypeScript interfaces (MonthlyMetrics, MonthlyAnalysis) to type the monthly analysis data structure.
• Updated chart imports to support new visualizations.

Evaluation Data Management Utilities

• Added scripts/clean_dynamic_eval.py, a utility script to safely delete dynamic evaluation data from BigQuery, with interactive and force modes, to prepare for unbiased (hourly) evaluation.
• Updated scripts/populate_dynamic_evaluation.py to default to hourly (--interval 1) predictions for unbiased evaluation, and added a warning if a larger interval is used, to prevent temporal bias in metrics. [1] [2]