Adjust ip ranges to reduce noise in random traning data

[Elsensee]

Because of the huge address space that IPv6 has, 7 out of 8 addresses will probably never be seen in real world login attempts which could reduce the effectiveness of the trained model.

I wondered if I should include the LLA fe80::/10 (well /64 really per spec), but decided against it, because these aren't the addresses we want to protect against as they are purely local and on a server setup very unlikely. (No router must be involved)
Also, I decided against ULA fc00::/7 because these would probably be more used within a company and then the other training approach of scrambling valid login attempts with other users' IP addresses would be the better method.
And so I focused only on 2000::/4 as this is the dangerous internet, as we currently know it.

Right now, this is totally untested because I am mainly looking for feedback for this approach. What do you think?

This would close #959

[github-actions]

Hello there,
Thank you so much for taking the time and effort to create a pull request to our Nextcloud project.

We hope that the review process is going smooth and is helpful for you. We want to ensure your pull request is reviewed to your satisfaction. If you have a moment, our community management team would very much appreciate your feedback on your experience with this PR review process.

Your feedback is valuable to us as we continuously strive to improve our community developer experience. Please take a moment to complete our short survey by clicking on the following link: https://cloud.nextcloud.com/apps/forms/s/i9Ago4EQRZ7TWxjfmeEpPkf6

Thank you for contributing to Nextcloud and we hope to hear from you soon!

(If you believe you should not receive this message, you can add yourself to the blocklist.)

[ChristophWurst]

Thanks a lot! I'll give this a test :)