feat: add image semantic and emotions analysis

[varshneydevansh]

Issue

The mush needed and the main feature of the extension is to able to

Thumbnail Analysis (Content Sensitivity):

The Challenge: This is even harder than semantic text analysis directly in the extension. Analyzing images for concepts like "sadness", "gore", or "unwanted content" requires sophisticated computer vision AI models. These are too large and computationally expensive to run effectively within the extension itself.

We could implement options to blur or completely hide thumbnails based on the existing keyword/channel filters. For example, if a video is hidden due to the keyword "sad", we could add an option to just blur its thumbnail instead of hiding the whole video block. This is much more achievable with CSS.