Properly handle unicode and grapheme clusters

[terror]

This PR addresses issues with Unicode and grapheme cluster handling in the markdown parsing and presentation logic.

To summarize the changes:

• Replaced character-based operations with grapheme-aware logic using the unicode-segmentation crate.
• Updated RopeExt trait to use byte-based operations for better Unicode compatibility.
• Modified the parser to correctly handle grapheme clusters in code block fence detection.
• Adjusted diff application to work with byte offsets instead of character offsets.
• Added integration tests for complex Unicode scenarios.