Feat/llm page tracking

[raymondcen]

This PR adds page-level traceability to extracted metrics from PDFs.

• Inserts [PAGE N] markers at the start of text from each PDF page during extraction.
• Updates metric extraction to detect these markers and record the page numbers where each metric appears.
• Adds a source_pages field to metric outputs.
• Improves logging to retain original text and report input size.
• Adds required imports to support page parsing logic.