FIP: Shannon entropy over purchase distribution weighting of customers

[reisepass]

We propose to add a customer weighting coefficient proportional to the shannon entropy of their purchase vector.

Customers that purchase from multiple different protoocls and providers are better customers and less likely to be a sybil. It is common in a marketplace setting for a customer to continue purchasing other services once they have gotten onboarded to purchase one. Customers that pay for one AI services are likely to be in the target group of another. If a customer finds Forest AI for its high accuracy in niche machine translation tasks they will see all other services, already have an account with funding and might then buy LLM services from another provider.