Coin mixes save vm on fail with regular workitems

When an integration or a status check is running and an another status check is triggered with save VM on failure from the same sha1, the VM that eventually fails will be saved for the user that the first workitem had.

Expected behaviour would be to save the VM under the user that scheduled with the save VM on failure flag, being first or last to schedule. If the items are not separated completely it could lead to issues when multiple save vm on failure are triggered from same sha1.