Fix SDE inference producing noise when shift != 1.0

The SDE inference path computes `next_timestep` using a linear formula:
```
next_timestep = 1.0 - (float(step_idx + 1) / infer_steps)
```
This ignores the shift transformation applied to the timestep schedule:
```
t = shift * t / (1 + (shift - 1) * t)
```
When shift != 1.0, the renoise step uses a timestep that doesn't match the shifted schedule the model expects, causing accumulated error and noise-only output.

`t_prev` is already available from the loop iterator and includes the shift transformation. The fix replaces the linear calculation with t_prev:

# Before:
```
next_timestep = 1.0 - (float(step_idx + 1) / infer_steps)
xt = self.renoise(pred_clean, next_timestep)
```

# After:
```
xt = self.renoise(pred_clean, t_prev)
```
When shift == 1.0, the behavior is identical to the current code. When shift != 1.0, the schedule is now correct.

modeling_acestep_v15_base.py

@@ -1941,8 +1941,7 @@ class AceStepConditionGenerationModel(AceStepPreTrainedModel):
                     # Stochastic Differential Equation: predict clean, then re-add noise
                     t_curr_bsz = t_curr * torch.ones((bsz,), device=device, dtype=dtype)
                     pred_clean = self.get_x0_from_noise(xt, vt, t_curr_bsz)
-                    next_timestep = 1.0 - (float(step_idx + 1) / infer_steps)
-                    xt = self.renoise(pred_clean, next_timestep)
+                    xt = self.renoise(pred_clean, t_prev)
                 elif infer_method == "ode":
                     # Ordinary Differential Equation: Euler method
                     # dx/dt = -v, so x_{t+1} = x_t - v_t * dt