archive download of a valid file might not work with all servers

## Details

Observed in https://github.com/nextcloud/app-certificate-requests/pull/316

- have a server that sends those headers when fetching the tarball

```
HTTP/1.1 200 OK
Date: Mon, 22 Jun 2020 21:14:25 GMT
Server: Apache/2.4.43 (Unix)
Upgrade: h2,h2c
Connection: Upgrade
Last-Modified: Mon, 22 Jun 2020 17:05:32 GMT
ETag: "4efea-5a8af401def00"
Accept-Ranges: bytes
Content-Length: 323562
Content-Type: application/x-tar
Content-Encoding: x-gzip
```
- try to publish a release
- see "not a gzip file" as error message
- any reproduction with curl+tar, but also with python shell fails
- error was sent out deeply in python's own gzip module when checking the magic numbers

I suppose that either the server did not serve the real file to the appstore client, but something else, or that there is a transport issues on the way (double-gzipped). Faking the python-user agent with curl still yielded the correct file.

It might still be the download server doing something funky. A clearer message (or at least logging the downloaded type, of not payload) could help to narrow it down.

## JavaScript Console Output

- 400 bad request, "not a gzip file"

## Steps to reproduce

(see above)