Page images can overwrite each other in the PDF export

1. Create a corpus.
2. Create a folder in that corpus.
3. Add two pages with different images, but with the same name.
4. Export that.
5. Run the PDF export on that export.
6. Get two duplicate images.
7. Cry.

The resulting PDF for the folder will include twice the image of the second page. The first page was saved in the temporary directory as {name}.jpg, but the second one has the same name so it overwrites it and only the image of the second page gets used.

You can also use that for evil purposes:

1. Create a corpus.
2. Create a folder in that corpus.
3. Add a random page named ../../../../../../../../../../../../lol
4. Export that.
5. Run the PDF export on that export.

6. PermissionError: [Errno 13] Permission denied: '/tmp/tmpxet51my3/../../../../../../../../../../../../lol.jpg'

7. Cry.

Note that since this is related to automatically generated filenames, this might also happen when generating the final PDFs (on folder names), or on ALTO exports if we implement #43.