docs: add wise-ft to clip finetuning

[bwanglzu]

In this PR, we start to support WiSE-FT fine-tuning for models with a strong zero-shot capability, to merge weights between pre-trained model and zero-shot model. Currently all CLIP-family models are supported. Functionality has been implemented inside core. To make use of it:

import finetuner

run = finetuner.fit(
     model='ViT-B-32#openai',
     ...,
     loss = 'CLIPLoss',
-    callbacks=[],
+    callbacks=[WiSEFTCallback(alpha=0.4)],
)

Review documentation change here

---

• This PR references an open issue
• I have added a line about this change to CHANGELOG

[github-actions]

📝 Docs are deployed on https://ft-docs-wise-ft--jina-docs.netlify.app 🎉