Create proof of concept of ImageTextLayer-based text layer in PDF.js

As a possible solution for https://github.com/hypothesis/client/issues/5023, explore creating a custom text layer in PDF.js that re-uses the `ImageTextLayer` class that we created for VitalSource PDF books.

**Goals:**

- Find out if using ImageTextLayer would enable us to improve text-selection behavior in PDF.js
- Find out if there are potential improvements in ImageTextLayer-based selection for VitalSource PDFs. A potential benefit of this approach is that we could end up with a single text layer implementation across all image-based content, which we can focus on polishing for different platforms etc.
- Get a better understanding of what makes text selection on iOS work well/poorly - ie. why does PDF.js's own text selection layer interact poorly with iOS Safari's text selection
- Measure performance of custom text layer vs base layer. TBD exactly what we measure and how.

The outcome should be a functional prototype behind a feature flag, some notes on what we learned about iOS text selection, and a decision about whether we want to use an ITL-based text layer on iOS and/or other platforms.