test_get_object_torrent: treat torrent body as bytes (not UTF-8)

In `s3tests/functional/test_s3.py`, `test_get_object_torrent()` calls `_get_body(response)` which unconditionally does `got.decode()` when the body is bytes.

A `.torrent` file is a bencoded *binary* format. The `pieces` field contains raw SHA1 digests, so the response body is not guaranteed to be valid UTF-8. When a backend returns an actual torrent body, the test can fail with `UnicodeDecodeError` even though the server is correct.

Suggested fix: in `test_get_object_torrent`, read `response["Body"].read()` as bytes and compare against `data.encode("utf-8")` (do not decode).

PR prepared on fork: drey123/s3-tests (branch omni-fix-torrent-bytes).