EV divergence check treats normal duty charging as an off-plan level #1244

The candidate compares measured EV progress with slot-average watts spread evenly through the interval. Duty dispatch intentionally charges at the allowed peak and stops after spending the energy budget.

Local reproduction: a 60 kWh EV, a one-hour action averaging 5.5 kW with an 11 kW peak. Normal delivery after 15 minutes exceeds the uniform estimate by more than two percentage points and causes a needless replan. The regression also covers an execution start halfway through the slot.

Expected: compare progress with the feasible earliest/latest placement of the planned pulse, while still detecting corrections outside that range. Fix and regression: #1240.