Implement ML "Bubble Zoom" feature (Kavita+?)

[fubarpie]

The open source android project Seeneva/seeneva-reader-android maintains a comic reader Android app which uses a ML object detection model based on YOLOv4 tiny to detect speech bubbles and allow users to zoom speech bubbles out for easy reading without manually zooming into panels and bubbles. This is essentially a clone of the "Bubble Zoom" feature that was introduced to Google Play Books in 2016, and is actually an excellent QOL feature that I use very often.

I propose that Kavita borrows this object detection model to implement the same feature in its reader.