retain thin provisioning support for volume migration

OpenStack Block Storage enables you to choose a volume back end
based on virtual capacities for thin provisioning using the oversubscription
ratio. In some cases, thin provisioning volume will become thick
provisioning volume after volume migration. Storage free capacity
will dramatically drop if a lot of volumes has been migrated.

Currently, there are three ways for volume migration.
- Cinder storage driver
- Cinder volume manager
- Nova compute
The first is the driver specific migration. Since some storage
back-ends have their special commands to copy the volume, Cinder
also provides a way for them to implement in terms of their own
internal commands to migrate. Whether volume is still thin
provisioning after volume migration depends on the specifc
drivers.
The second is the generic host-assisted migration, which
consists of two different transfer modes, block-based and file-based.
The block-based transfer mode is done by 'dd' command which
supports sparse on copy and thus this mode might retain
thin provisioning support after volume migration for available
volume such as LVM driver. The file-based transfer mode is done
by file copy and this mode cannot retain thin provisioning support
after volume migration although volume status is available. The latter
typically applys to the RBD driver.
The nova compute will start a block job for volume migration for in-use
volume and thin provisioning volume must become thick provisioning
volume after volume migration.

This introduces some methods making sure thin provisioning volume
will reatin thin provisioning identity after migration despite what migration
ways.