n-cpu attempts to disconnect volumes after early pre_live_migration failures on the destination during a live migration

Description
===========

When live migrating an instance with volumes attached pre_live_migration on the destination will initially attempt to map these volumes to the destination by creating a volume attachment (cinderv3) or calling initialize_connection (cinderv2).

At present if either call fails the generic live migration rollback code is called and an attempt is made to disconnect volumes from the destination ignoring the fact that they have not been mapped or connected to by that host.

Steps to reproduce
==================

* Live migrate an instance with volumes attached, ensuring calls to either cinder API fail during pre_live_migration.

Expected result
===============
No attempt is made to disconnect volumes from the destination.

Actual result
=============
n-cpu attempts and fails to disconnect volumes from the destination using connection_info from the source bdms.

Environment
===========
1. Exact version of OpenStack you are running. See the following
  list for all releases: http://docs.openstack.org/releases/

117508129461436e13c148bb068b0775d67e85d3

2. Which hypervisor did you use?
   (For example: Libvirt + KVM, Libvirt + XEN, Hyper-V, PowerKVM, ...)
   What's the version of that?

Libvirt

2. Which storage type did you use?
   (For example: Ceph, LVM, GPFS, ...)
   What's the version of that?

N/A

3. Which networking type did you use?
   (For example: nova-network, Neutron with OpenVSwitch, ...)

N/A

Logs & Configs
==============