Fix UTF-8 decoding error for multi-byte sequence at chunk boundary

[k-lukas]

Summary

HTTP chunked transfer encoding can split multi-byte UTF-8 characters (like box-drawing characters ─, ┴) across chunk boundaries. The previous code decoded each chunk independently, which caused spurious "Invalid UTF-8 in response" errors on perfectly valid output — particularly with server-side formats like PSQL that use box-drawing characters in table borders.

This PR:

• Extracts a ChunkDecoder struct that carries incomplete multi-byte sequences across chunk boundaries, only reporting an error when bytes are genuinely invalid
• Improves the error message for real UTF-8 errors to include the byte position, a hex dump around the bad bytes, and a text preview — making it much easier to diagnose whether the issue is corrupted data, wrong encoding, or unexpected binary content
• Adds 11 unit tests covering split 2/3/4-byte characters, byte-at-a-time feeding, truncated streams, invalid bytes, and global position tracking

[k-lukas]

@cursor review

[cursor]

Skipping Bugbot: Bugbot is disabled for this repository. Visit the Bugbot dashboard to update your settings.