Fix UTF-8 char boundary handling in cursor movement and rendering

[jvanderberg]

Summary

This PR fixes potential panics and incorrect cursor positioning when working with multibyte UTF-8 characters. The changes ensure that cursor positions are always snapped to valid UTF-8 character boundaries during vertical movement and rendering.

Key Changes

• Buffer vertical movement: Added snap_to_char_boundary() helper method that walks backwards from a byte offset to find the nearest valid UTF-8 character boundary. This method is now called in move_up() and move_down() to ensure the cursor never lands in the middle of a multibyte character.

• Cursor rendering: Updated render_editor() in the UI layer to snap the cursor column to a valid char boundary before rendering. Also fixed the cursor character extraction to properly handle multibyte characters by using chars().next() instead of byte-based slicing, which could panic or produce incorrect output on non-ASCII text.

• Parser safety: Added bounds checking in fixup_link_lines() to use .get() instead of direct indexing, preventing potential panics on edge cases.

Notable Implementation Details

• The snap_to_char_boundary() method uses line.is_char_boundary() to safely identify valid UTF-8 boundaries, walking backwards one byte at a time until a boundary is found.
• Cursor character rendering now correctly handles multibyte characters by extracting the full character width using chars().next() and to_string(), then calculating the proper offset for the "after" portion.
• Comprehensive test coverage added for both move_up() and move_down() with multibyte UTF-8 characters (Greek letters) to prevent regression.

https://claude.ai/code/session_01J88TRUMkjBFCGXeBwcMNSk