AI Adjustment: crowd-sourced corrections

Actual AI strength at the same numeric AI level is known to vary across classes / cars. We could crowd-source collecting adjustments that'd get transparently applied to the recommended AI levels.

Ideally this would be transparent, data would be collected through the app (which assumes that some backend service exists, unlike today)

The issue describes a backend service to collect and apply AI strength adjustments across car classes. Start by examining the existing codebase for AI level recommendations and data structures. Look for where AI levels are currently set and how car/class data is handled. Determine how to transparently apply crowd-sourced corrections, likely requiring new API endpoints and data storage.

索引モデルが issue の本文から書いたものです。