[FEATURE] Add image search pretrained models

[IanMenendez]

Currently, there are no image search pretrained models in OS

Having image search pretrained models would make it plug-and-play removing unnecessary friction.

It would be good to have openai clip models

The workaround to this issue is one needs to trace the model themselves and then register it

[IanMenendez]

Would be also cool to have some pretrained multimodal search models
Something like visual bert

I've been trying to setup this multimodal model but the lack of documentation makes it really hard

[mingshl]

tagging to the RFC [RFC] Support more local model types

[IanMenendez]

Right now ml commons does not support image embedding models so it is impossible to add image embedding models as pretrained models

To support this we need to add image embedding model type: #2622

We will also need to add an image embedding ingest processor to enable the ingest of images into image embedding models

[hijakk]

+1 for interest in native CLIP support in OpenSearch