Avoid client-side file upload to bucket; use server-side

[j-mao]

File upload to bucket is used for processing submissions and resumes. Currently, in both cases the API returns a signed URL which the client can then use to upload the file directly to the bucket.

However if the client goes wrong (which internet browsers are wont to do), then the backend database records an upload, but the upload does not occur successfully. Specific causes may include the end-user refreshing the page before the upload completes, or unexpected errors during upload (e.g. CORS). This may be a cause of #194.

Using server-side file uploads will better ensure that anything recorded in the database actually exists. Additionally we can do the same thing to implement #24.