Refactor code block syntax checker for maintainability and clarity

[theotrosman]

This script (check_code_blocks.py) is useful for validating embedded Python code in Markdown files using pyright, but it could benefit from a few improvements in structure and clarity to support better maintainability and robustness.

Observations:
Code block extraction logic is coupled to Python-only blocks.
The script hardcodes ```python as the start delimiter. If in the future we want to support other languages or extensions, this makes it harder to scale.

The ignore-check logic is verbose and nested.
The nested all(all(...)) logic for checking imports from specific modules (autogen_*) is functional but hard to read and could be abstracted into a helper function.

Subprocess call to pyright is inline.
This can be abstracted for better testability and reusability.

Temporary files are written and not cleaned up.
Currently, delete=False is used in NamedTemporaryFile, but there's no explicit cleanup. This may leave behind orphaned .py files after execution.

No test mode or dry-run flag.
Could be useful for CI pipelines to add a non-blocking mode that just reports without throwing RuntimeError.

Hardcoded dependency on pyright.
There's no fallback or clear error message if pyright is not installed.

Suggested improvements:
Abstract extract_python_code_blocks() into a more generic extract_code_blocks(language: str)

Refactor the should_ignore_block() logic into its own function for readability

Wrap the subprocess call in a run_pyright(file_path: str) -> CompletedProcess utility

Add optional cleanup for temporary files

Add a --dry-run CLI option that doesn’t raise errors (useful for CI)

Check for pyright presence at startup and fail fast with a helpful message if not found