Fix itrex WOQ model loading

[echarlaix]

Fix itrex WOQ model saving : currently the quantization_config is not saved when calling save_pretrained, resulting in issues when trying to load the model after saving it

also enable the possibility to :

• load a safetensors checkpoints
• provide a subfolder argument

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.