Panic when truncating multibyte session text #3

Summary

csx can panic when a session title or preview contains multibyte UTF-8 characters, such as Korean text.

Reproduction

Run csx with a Codex session whose title/preview contains Korean text long enough to be truncated in the picker row.

Actual behavior

The process panics while formatting the row:

thread 'main' panicked at src/formatting.rs:15:28:
end byte index 157 is not a char boundary; it is inside '그' (...)

Expected behavior

Session text should be truncated without panicking, preserving valid UTF-8.

Root cause

compress_text() uses byte length and byte slicing (clean.len() and &clean[..max_length - 3]) even though Rust string slices must be on UTF-8 character boundaries.

Environment

• Agent Session Hub: v0.3.6
• OS: Ubuntu Linux aarch64
• Provider: Codex