Fix Python 3 and modern OpenCV compatibility

[RBQAAQ]

Summary

Fix compatibility issues in Howdy 2.x when running with Python 3 and modern OpenCV/NumPy versions.

Changes

• Support both Python 3 configparser and the legacy Python 2 ConfigParser module.
• Replace deprecated datetime.datetime.utcnow() with a timezone-aware UTC datetime.
• Flatten the OpenCV histogram result before summing and iterating over it, preventing scalar indexing errors with modern OpenCV and NumPy.

Problems fixed

Before these changes, I encountered:

ModuleNotFoundError: No module named 'ConfigParser'

IndexError: invalid index to scalar variable

DeprecationWarning: datetime.datetime.utcnow() is deprecated

Tested environment

• Arch Linux
• Howdy 2.6.1
• Python 3.14
• OpenCV 5.0.0
• pam_python3.so
• UVC infrared camera using an 8-bit GREY stream

Validation

I verified that:

• howdy test displays the infrared camera feed correctly.
• howdy add successfully enrolls a face model.
• PAM authentication through sudo succeeds.
• Password fallback works after face detection times out.
• The modified files pass python3 -m py_compile.

This branch is based on the v2.6.1 tag. GitHub reports that it cannot be merged automatically into dev, so the commit may need to be rebased or cherry-picked onto the appropriate branch.

[RBQAAQ]

Closing this pull request because two parts overlap with existing pull requests:

• fix: make test command histogram logic shape-agnostic for OpenCV 5 #1128 already contains the OpenCV 5 histogram compatibility fix.
• Add Python 3 PAM compatibility #848 already contains the Python 3 ConfigParser compatibility fix.

The remaining datetime change is already fixed on the current mainline, while #1111 concerns a backport to Howdy 2.6.1. Since there is currently no dedicated 2.x maintenance branch, this PR cannot target an appropriate merge base.

I will add my real-system validation results and tested commit to #1111 so maintainers can cherry-pick or use it for a future 2.6.1 maintenance release.