Fix full-width character alignment

[Selina-0017]

Summary

Fix incorrect caret alignment in error context rendering when input contains full-width Unicode characters (e.g. Chinese characters).

Previously, caret positioning used character length (len()), which does not reflect terminal display width for full-width characters.

This caused misaligned error indicators in UnexpectedInput.get_context().

Changes

• Replace naive character-count spacing with display-width-aware computation
• Add regression test for full-width Unicode input

Testing

• Added test covering full-width character alignment
• All existing tests pass

Notes

No external dependencies added. Uses Python standard library only.

[erezsh]

Hi, I ran it through a reviewer, can you address this finding?

Tab + full-width lines now overshoot the caret (a small regression). expandtabs() computes tab stops by character count while _display_width counts display columns, and the two compose
incorrectly. Verified: for before = "你好\tX", a terminal renders the error at display column 9, the old code correctly emitted 9 spaces, and the new code emits 11. So for lines mixing wide chars and tabs, the PR
makes alignment worse than before. Fixing it properly needs a width-aware tab expansion (advance to the next multiple of 8 in display columns) instead of _display_width(before.expandtabs()).