VLMEvalKit support for nanoVLM

[MaoSong2022]

I can help to integrate nanoVLM into VLMEvalKit.

the goal is to enable evaluating the model built upon nanoVLM with benchmarks supported by VLMEvalKit.

ideally, this leads to the following format:

# Load pretrained weights from Hub
from models.vision_language_model import VisionLanguageModel

model = VisionLanguageModel.from_pretrained("lusxvr/nanoVLM-222M")

# evalutation
cd path/to/VLMEvalKit/
python run.py --model lusxvr/nanoVLM-222M --data MME 

[lusxvr]

This would be greatly appreciated! We have an internal/hack-y way to do it by porting the weights into the SmolVLM architecture (#74) but if you integrate it into VLMEvalKit directly this would make it easier!