fix(parser): page_count uses fitz total but pages tuple is shorter after per-page skip -- section page_end out of range #288

Summary

Fix #275 added per-page try/except so corrupt pages are skipped with a warning. But ParsedDocument is constructed with page_count=doc.page_count (total fitz page count including skipped pages) while pages=tuple(pages_list) only contains successfully parsed pages. Downstream code (toc_extractor._derive_page_ranges) uses page_count as the ceiling for the last section's page_end, creating section ranges that reference page numbers with no corresponding Page object in doc.pages. This causes silent gaps in chunking where chunks are expected but never created.

Fix

ingestion/parser.py after the per-page loop: set page_count=len(pages_list) so it matches the actual number of parsed pages.

Impact

Documents with corrupt/skipped pages have incorrect TOC section page ranges. Chunks expected for those sections are never created, producing silent content gaps.