[wwb] Add possibility to check video inputs for VLM

[sbalandi]

Description

WWB is updated to run VLM pipeline with video inputs.

• New model type has been added for this purpose, so to configure the video input it's needed to run wwb with --model-type visual-video-text.
• Default video data is data from dataset: lmms-lab/LLaVA-Video-178K , subset 30_60_s_academic_v0_1 , video from archive 30_60_s_academic_v0_1_videos_10.tar.gz (was chosen because of its weight - 274 MB). Archive includs 56 videos from different datasets: youcook2, NextQA, ego4d, Chrades and activitynet.

how to use:
optimum-cli export openvino -m Qwen/Qwen2-VL-7B-Instruct --weight-format int8 qwen2-vl-7b-Instruct
python whowhatbench/wwb.py --base-model qwen2-vl-7b-Instruct --model-type visual-video-text --gt-data vlm_video_gt.csv

Ticket:
CVS-173847

Checklist:

• Tests have been updated or added to cover the new code.
• This patch fully addresses the ticket.
• I have made corresponding changes to the documentation.