Plugin: Enable exact image/video matches using perceptual hashes

Implement a perceptual hashing plugin using the Thorn Perception library (feel free to use a different library if needed) to detect duplicate and near-duplicate media in forensic investigations. The library supports various hashing algorithms. Start by supporting a few or even a single hash and support more if possible.

Key operations:

1. Given a directory of images/videos, create a database of hashes for all the images/videos in the directory (similar to the face-match plugin).
2. Given a directory of images/videos, and a database of hashes (generated from the above endpoint), check each image/video against the database of hashes. For each image/video, return zero or more images/videos from the database that match the query image/video.
3. Support import and export of hashed databases.

Optimize performance after the initial implementation is complete.

Start by reading the existing face-match plugin, then review the Thorn Perception library and its supported hashing algorithms. Implement the three described operations: build hashes for directory media, query a hash database for matching media, and import or export hashed databases. Done means the plugin supports the initial hash workflow for images or videos and returns matching media; performance optimization follows afterward.