Add MedCLIP Model for Zero-shot Inference

[wuzijian1997]

Add a new model medclip_zeroshot.py

Note that the original MedCLIP repository seems to have bugs. I mainly modify the "MedCLIPFeatureExtractor" class. The source code lacks some image transformation functions, so passing the image transformation will get errors. I kindly remind you to fix this issue for the original repository.

[wuzijian1997]

I add an example of the MedCLIP zero-shot classification. Please check it.