fix: IpynbConverter.accepts() catches UnicodeDecodeError on non-ASCII files (fixes #1894)

[hanhan761]

Summary

\IpynbConverter.accepts()\ reads the full file stream and decodes it to detect Jupyter notebooks. When a non-ASCII file (e.g., a French PDF with accented characters like é/è/à) produces a MIME type starting with \�pplication/json\, the \.decode()\ call raises \UnicodeDecodeError\ which propagates uncaught and crashes the entire conversion pipeline.

This fix wraps the decode block in \ ry/except (UnicodeDecodeError, ValueError)\ so that \�ccepts()\ gracefully returns \False\ instead of crashing.

Issue

Fixes #1894

Verification

• Binary non-ASCII content with \�pplication/json\ MIME → returns \False\ (no crash)
• Valid \.ipynb\ notebook → still returns \True\
• Plain JSON → returns \False\
• 2/3 existing tests pass (1 pre-existing failure due to missing docx dependency)

[Adam Fourney (afourney)]

Closed in #2360