[Bug]: Unload SD checkpoint to RAM not working

### Checklist

- [X] The issue exists after disabling all extensions
- [X] The issue exists on a clean installation of webui
- [ ] The issue is caused by an extension, but I believe it is caused by a bug in the webui
- [X] The issue exists in the current version of the webui
- [X] The issue has not been reported before recently
- [ ] The issue has been reported before but has not been fixed yet

### What happened?

"Settings/Actions/Unload SD checkpoint to RAM" cause no effect.
Complemmetar API call sdapi/v1/unload-checkpoint not working too.

In the original A1111 this works and free VRAM.

### Steps to reproduce the problem

1. Generate some image
2. note VRAM used by running nvidia-smi
3. Go to Settings/Actions, press "Unload SD checkpoint to RAM"
4. Rerun nvidia-smi: VRAM still filled.

### What should have happened?

Expected: VRAM mostly cleared after "Unload..."

### What browsers do you use to access the UI ?

Apple Safari

### Sysinfo

[sysinfo-2024-05-14-11-01.json](https://github.com/lllyasviel/stable-diffusion-webui-forge/files/15307436/sysinfo-2024-05-14-11-01.json)

### Console logs

```Shell
Begin to load 1 model
[Memory Management] Current Free GPU Memory (MB) = 14367.57470703125
[Memory Management] Model Memory (MB) = 159.55708122253418
[Memory Management] Minimal Inference Memory (MB) = 1024.0
[Memory Management] Estimated Remaining GPU Memory (MB) = 13184.017625808716
Moving model(s) has taken 0.03 seconds
Total progress: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 20/20 [00:05<00:00, 3.63it/s]
Total progress: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 20/20 [00:05<00:00, 3.97it/s]
```

### Additional information

_No response_

Reproduce the issue through Settings/Actions and the sdapi/v1/unload-checkpoint endpoint while monitoring VRAM with nvidia-smi. Trace the handler for that action and compare its behavior with the original A1111 implementation; done means the checkpoint unloads and VRAM is mostly cleared.

Written by the indexing model from the issue text.