Don't retry on 404s.

[escapewindow]

I noticed that we were retrying 404s for downloading the currently largely nonexistent chain-of-trust.json{,.sig} files in #294... this means it takes us a few minutes to realize that yes, the file is in fact missing. I think not retrying on 404 is ok; if the file is there but we have other issues, it'll be a non-404 status.

If we think this might have larger repercussions, I can retry on BaseDownloadError in most cases, and override that with just DownloadError specifically for downloading chain-of-trust.json{,.sig} until we create those files across all 3 worker implementations.

What do you think, @JohanLorenzo , @tomprince ?

[coveralls]

Coverage remained the same at 100.0% when pulling 2f72b66 on escapewindow:Download404 into 276f4a0 on mozilla-releng:master.

[escapewindow]

(This mainly works because this line doesn't include Download404 in the list of exceptions to retry on).

[JohanLorenzo]

This approach makes sense to me. On top of my head, I can't think of a case where we want to wait on a 404. That might happen when an artifact isn't present in a given region. However, as this case occurs less often than failing on a missing CoT artifact, I think it's worth giving a try.

[escapewindow]

Let's give it a try.