Treat each linear spine document as a single reading unit instead of splitting one spine item into multiple chapters based on section heuristics.
Improve XHTML text extraction so inline spacing is preserved, list and container text is included in reading order, and non-readable tags are ignored before speech output.
Resolve TOC hrefs against the nav or NCX document location so chapter titles match nested content paths correctly, and skip non-linear spine items.
Add a regression test that builds a minimal EPUB fixture and verifies spine order, title resolution, preserved inline spacing, extraction of non-paragraph content, and exclusion of linear=no content.
Verified with: python test_epub_parser.py; python check_naming.py src/epub_parser.py; python check_naming.py test_epub_parser.py
