feat: enhance LLMClient with structured output capabilities

feat: enhance app.py with session state management and extraction strategy selection

feat: enhance architecture and developer documentation for clarity and detail

feat: enhance field and unique indices descriptions handling in app and agents

feat: update unique indices combinator to return array of objects

Refactor FieldMapperAgent to process entire documents at once, improving efficiency by eliminating page-by-page analysis. Update logging to reflect changes in document processing. Enhance AzureDIService to save extracted content to timestamped files for better traceability.

Refactor AzureDIService to enhance document analysis logging and update table extraction logic. Temporarily disable table extraction and improve content type logging. Update TableAgent to reflect changes in context handling.

Add cost tracking functionality across various components, including Executor, Planner, and FieldMapperAgent. Integrate CostTracker to monitor LLM and document intelligence costs, enhancing logging for cost-related metrics and providing detailed cost breakdowns in the user interface.

Revise documentation in app.py for Deep‑Research PDF Field Extractor, enhancing clarity on system architecture, core components, processing pipeline, and key features. Update usage instructions and support resources for better user guidance.

Update requirements.txt to pin versions of dependencies for consistency and stability, including altair, pandas, streamlit, pyyaml, python-dotenv, openai, pydantic-settings, PyMuPDF, and azure-ai-documentintelligence.

Update Dockerfile to use new app entry point and enhance requirements.txt with additional dependencies. Remove obsolete streamlit_app.py file.