fix: handle async client and None reward values

fix: correct StepResult import path

feat: create inference.py baseline script