Switch to using fully unweighted significant sales algorithm

[jeancochrane]

It seems like the fully unweighted significant sales algorithm (model run ID 2025-10-18-blissful-sean) performs better than the version that is currently deployed to prod reports, which ignores trees that increase error for training observations. Let's run the fully unweighted algorithm on the 2024 model and then deploy both 2024 and 2025 to HomeVal.

Running the algorithm on the 2024 model will require branching off of this commit in the res model, then making a similar set of changes as I did in this branch to bring the code up to date. It's kind of an annoying process, so let me know if you need any help.

[wagnerlmichael]

Is this still a priority? @jeancochrane I remember we talked about the complexity involved in this but can't remember what we ended up deciding

[jeancochrane]

@wagnerlmichael Let's check in with @ccao-jardine next week to determine whether this is a priority. I just bumped the ticket into next sprint as a reminder to discuss it.