#1965 fix(bench): prevent UnicodeDecodeError on Windows

Description

## Summary

Fix a `UnicodeDecodeError` in `BENCH.bat` on Windows systems using non-UTF-8 default encodings, such as CP932 (Japanese Windows).

## Problem

On Japanese Windows with Python 3.11, running `BENCH.bat` fails during machine information collection because `project_version()` reads `CMakeLists.txt` without specifying an encoding.

Python defaults to CP932, which cannot decode the UTF-8 BOM at the beginning of the file.

```text
UnicodeDecodeError: 'cp932' codec can't decode byte 0xef in position 0: illegal multibyte sequence
```

## Fix

Explicitly use `encoding="utf-8-sig"` when reading `CMakeLists.txt`.

This handles UTF-8 files with or without a BOM, regardless of the system's default encoding.

## Testing

- Reproduced the issue on Japanese Windows with Python 3.11.
- Confirmed that setting `PYTHONUTF8=1` works around the issue.
- [x] Verify that `BENCH.bat` completes without `PYTHONUTF8` after applying the fix.