Split import and processing pipeline

When a user uploads a file:

1. We save the file to the disk, create a Scan entry, and start the importing task.
2. In the importing task, we extract every image we can find in the file. We store this image information the Image table, containing: path, origin, status and scan_id.
3. After the import task is finished, we start the processing task for the scan_id.
4. The processing task reads all Image entries from the database, and processes them according to status, creating Page entries in the process.

The Image table:

• id: primary key
• path: path to the image on disk
• origin: path to the image in the original scan file
• status: one of normal, raw, failed.
• scan_id: points to the scan

The Page table:

• Add image_id
• Remove path

We can deduce what to do with an Image the following way (related to #455):

• The status normal/raw without a Page entry means: waiting for processing with the normal/raw pipeline.
• The status normal with a Page entry means: processed successfully with the normal pipeline, immutable.
• The status raw with a Page entry means: processed successfully with the raw pipeline, we still allow assigning a different student, page and copy manually.
• The status failed without a Page entry indicates the user can either manually assign student, page and copy, or delete the image altogether.