potential regression in openssl digests

[Casey Bodley <cbodley@xxxxxxxxxx>]

There was a regression in some rgw etag calculations after the switch to 
OpenSSL for cryptographic hashes. In this case, the bug in 
https://tracker.ceph.com/issues/39160 is that rgw was calling 
hash.Final() twice on the same object. This apparently worked fine with 
NSS, but the second call returns a different result with OpenSSL. Just 
sharing in case this pops up elsewhere!

Casey