fix(ipynb): handle string and null cell sources in IpynbConverter

[hsusul]

Summary

• Normalizes Jupyter notebook cell source field representations (str, list[str], None, or missing) in IpynbConverter.
• Enables correct line-by-line title extraction when a notebook cell source is a single string.
• Prevents TypeError and conversion failure when cell sources are None or missing.

Problem

According to the Jupyter notebook specification (nbformat v4), cell source fields may be represented either as a list of strings or as a single string (or None).
When source is a string (e.g., "# Title\nSome markdown"), IpynbConverter._convert() previously iterated over source_lines character-by-character ('#', ' ', ...). As a result, heading detection (line.startswith("# ")) never matched, failing to extract the document title.
Additionally, if a cell source is None or missing, "".join(None) raised a TypeError wrapped in a FileConversionException, causing MarkItDown to crash or fall back to raw plain text output.
Furthermore, if metadata in the notebook JSON is null, notebook_content.get("metadata", {}).get("title", title) raised an AttributeError.

Root cause

IpynbConverter._convert() assumed cell.get("source", []) is always a list of line strings, without normalizing string or None values.

Fix

• Added _get_source_lines(source: Any) -> list[str] helper method to normalize cell sources (str, list, None) into a list of line strings with preserved line endings.
• Used _get_source_lines() across all cell types (markdown, code, raw).
• Updated metadata dict access to safely handle metadata: null.
• Added unit test test_ipynb_converter_string_and_none_source covering string, list, and None cell sources.

Testing

• pytest packages/markitdown/tests/test_module_misc.py -k test_ipynb_converter_string_and_none_source — PASS
• pytest packages/markitdown/tests/test_module_vectors.py — PASS (109 passed)
• git diff --check — PASS (clean, no trailing whitespace)

Compatibility

100% backward compatible. All existing tests pass.