Set content-type on uploaded S3 blobs

Not sure if this is something I can enforce on the server, or if the client needs to do this, but I noticed that for a file name like `e.<random hex id>`, S3 shows the file type as `<random hex id>`. This may be harmless, but if we can tell AWS those files are arbitrary binary data, we should.
