We should be computing/comparing Azure content MD5 hashes

When downloading data from Azure, we can optionally request a content MD5; we can use this to verify that we downloaded the expected data.

When uploading data/staging blocks for a multipart upload we can compute an MD5 hash which will be used to ensure that we uploaded the correct data (a 400 HTTP response will be returned if the hash mismatches).

These are optional features in the SDK, however, I think we should be making us of these to possible error scenarios.