Make the storage layer more resilent to failures

When a storage failure happens the internally maintained
flow detail or atom details should not be updated; to ensure
this happens update a cloned atom/flow detail first and only
if the save succeeds do we update the internal version (and
perform any futher work).

Change-Id: Ib8da476b6be2c71b663e0fc659b5df24e6b9f095

Author: Joshua Harlow