New features

• New unpronounceable heading detection – Detects empty or meaningless headings (e.g., <h2>?</h2>) that lack accessible name content.
• Expanded PDF link detection – Added /pdf/ to PDF source patterns for more comprehensive detection of PDF-linked content.
• Enhanced suspicious alt text detection – Added "portrait" to the suspicious alt word list. (#158, thanks @brianteeman)

Bug fixes

• Fixed alt text validation for CJK text – Unicode scripts (Chinese, Japanese, Korean) are now properly validated, eliminating false positives for space-less word structures and full-width punctuation.
• Fixed link validation false positives – Links with aria-label and decorative images no longer trigger incorrect warnings. (#159, thanks @matzesa)
• Fixed PDF link detection – Corrected invalid link sources preventing QA_PDF test from executing (#157, thanks @itmaybejj)
• Restored PUA characters in accessible name computation – Private Use Area (PUA) characters are no longer incorrectly removed during name calculation.

Thanks to @Michael-Nesci for finding bugs and testing!