Extend spam detection to low-effort issues

I'm thinking we could start being more aggressive with spam detection. Issues without a single word, containing only screenshots or logs, do not provide any reasonable value and should probably be treated as spam too.

I'll collect a few examples here for testing and then start with the implementation if you agree.

Examples:
https://github.com/nuxt/nuxt/issues/29946

No files, tests, or entry points are identified in the issue. Review the existing spam-detection implementation and the linked Nuxt example first, then define tests for issues with no words and issues containing only screenshots or logs; done means these cases are consistently classified as spam.

Written by the indexing model from the issue text.