I think your overshoot is caused by having too much integral and too little proportional action.
Integral action should only happen in steady state, but you already have quite a bit of integral windup after the first hour of the chart. This causes the fridge setpoint to stay too low for too long.
The integral is held to zero as long as the fridge setpoint is clipping to min or max.
I would decrease the maximum difference between beer and fridge setpoint.
I would also increase Ti, so it builds up slower. This is the beer to fridge Ti.
Also update to 0.5.x. Some anti-windup improvements were part of the 0.5.0 update.
Your heater can probably have a higher Kp. It is now only 10% on for a 1 degree difference.