upload script

Uploading PDFs is dangerous and error-prone, whether dealing with bulk uploads after a conference ingestion or individual papers after processing a revision. Files get put in the wrong place; sometimes we forget to upload them; and old files can overwrite new ones. I would like to move away from our current direct rsync/scp approach in favor of a script which applies some checks.

Here is the CLI I envision:

```bash
upload.py [file list]
```

The script will process all files and do the following for each:

* Infer its location on the server (see below)
* Ensure that the file is listed in the XML
* Ensure the file’s checksum matches the checksum listed in the XML
* If all checks pass, put the file in place via scp

This should be written so that the `upload()` function is available via the API, so that other scripts can use it directly.

Inferring the file location is easy. Here are some example file names:

* `2021.acl-1.pdf` → `$ROOT/pdf/acl/`
* `P19-1004.pdf`→ `$ROOT/pdf/P/P19/`
* `W19-6319v2.pdf` → `$ROOT/pdf/W/W19`
* `2020.emnlp-main.4.Attachment.pdf` → `$ROOT/attachments/emnlp`
* `2021.acl-main.4.mp4` → `$ROOT/videos/acl`
* `N13-2014.mp4` → `$ROOT/videos/N/`