V4.0.0 Custom ONNX Runtime Model

• Added custom-trained ONNX captcha model as the new default OCR method (~98% accuracy)
• New OCR priority: onnx > ddddocr > captchacracker > easyocr
• ONNX method requires only onnxruntime, numpy, and pillow — no opencv-python needed
• Improved browser header spoofing with randomized Chrome versions and proper sec-* headers
• Model files included in model/ directory (captcha_model.onnx, captcha_model_metadata.json)

3.0.0 OCR Options

• Refactored, streamlined OCR calling logic, updated summary statistics.
• Added support for ddddocr (default/recommended) and captchacracker alongside easyocr. Select via --ocr-method.
• Replaced --all-images with --save-images [0|1|2|3] for more control over saving successful/failed captchas to captcha_images folder.
• Successfully validated codes will be saved as <code>.png and unsuccessful ones as FAIL_<fid>_<timestamp>.png depending on --save-images settings.
• Added colored console output (requires colorama), consistent Nickname (FID) prefix, specific log levels.

2.4.5 EasyOCR Release

This is the final version that uses only EasyOCR as the Captcha solving method.