Keep screenshot on operation result page for further upload to Penguin Statistics.

[YuiTH]

Each time we call get_scene and it's result is ope_finish, we could call save_screenshot with new subdir name drop or something.

After that, simply helping users to upload their screenshots to Penguin Statistics is enough. We don't have to implement item recognition by ourselves, because Penguin Statistics already has an elegant classify model available in any browser.

We may need:

1. A new switch in config.yaml.
2. A new argument could be parse into operation mode.
3. Some limit to help users deal with screenshots (warning when close to limit).

[ZhaoZuohong]

或许可以利用 wasmer-python 之类的库，把企鹅物流的识别在本地运行