fix(ipynb): preserve JSON documents that mention notebook fields

[Oscar (Oscar-Williams)]

An ordinary JSON document such as {"description": "nbformat_minor describes a notebook version"} currently converts to empty Markdown. IpynbConverter.accepts() matches notebook field names anywhere in the text, then conversion treats the missing cells as an empty notebook and prevents the plain-text fallback.

Probe the parsed top-level JSON fields before accepting it as a notebook. Check JSON before trusting a .ipynb extension, since format detection can infer that extension for JSON schema fragments too. Preserve BOM/charset handling and stream position, and decline malformed or excessively nested JSON without interrupting conversion.

Validation on Windows, Python 3.11.7:

• The same 17 new regression cases give 12 failed / 5 passed on main and 17 passed with this fix.
• hatch test --python 3.11 -q --tb=short: 859 passed, 34 skipped, using the project's CI skip mode (GITHUB_ACTIONS=true, PYTHONUTF8=1).
• pre-commit run --all-files and git diff --check: passed.

Implemented and tested with Codex assistance.

[Oscar (Oscar-Williams)]

@microsoft-github-policy-service agree