VLM Support for vLLM Deployment

[movchan74]

Summary:
This update adds support for Vision-Language Models (VLM) to vLLM deployment, along with necessary tests and documentation updates. Only images are supported at the moment. A bug with insufficient resources error has also been addressed.

Key Changes:

1. VLM Support for Images: Added Vision-Language Model (VLM) support for vLLM deployment, currently limited to image processing.
2. Tests for VLM: Introduced tests for VLM on images to ensure proper functionality in vLLM deployments.
3. Documentation Update: Updated the vLLM deployment documentation to reflect new features and usage details.
4. Bug Fix: Resolved an issue where deployments could fail due to insufficient resource error because it was restarted.

Note:
Currently, there is no resizing for images in the deployment since vLLM doesn't allow passing arguments into the image processor (see vLLM Issue #7861). It seems the vLLM team is working on this. If needed, we can implement our own preprocessor for image resizing or resize the images before feeding them into the vLLM deployment.

Closes: #170

[appoose]

This is fine, could reproduce results ( have not checked complex videos though ). But let us merge it.