make scalar function choice weights dynamic and computable

[WiggidyW]

currently in the alpha_function branch, it handles scalar function choices by making the earlier ones have a lower score, and the later ones have a higher score

the original intent behind making choice weights fixed is because they represent a bad -> good range

with the computed profile being only weights across agents and across function tasks, the profile is human-readable. because, simply looking at the weights for different agent personalities and for different tasks gives you a lot of information that isn't very hard to interpret.

for example, if the computed profile is for a function that handles subjective preferences, then you could learn about the person it's simulating from the profile.

if we made these weights dynamic, computed profiles suddenly become non-human-readable.

but, now that agents can have access to MCP tools (in the alpha_function branch), and in the future will sometimes be able to communicate with each other (#136), the weights are gonna be pretty impossible to interpret anyways.

so, just focus on making profile computation as powerful as it can be, and focus on making the simulation as accurate as it can be.

make scalar function choice weights dynamic and computable, part of the profile