[FEAT] Support OpenVINO vision model

[szeyu]

Is your feature request related to a problem? Please describe.

Reference :

• OpenVINO Phi-3-vision notebook
• ov_phi3_vision.py

OpenVINO has created its own python code ov_phi3_vision.py to support vision model

Describe the solution you'd like

Vision Langauge Model Support:

Quantised vision model weight

Model	Model Link
Phi-3.5-vision-instruct-int4-ov	link

Additional context

Vision model and Non-Vision model should be initialised in a separated way as both model loaded in different way.