Refactor volume migration.

Split the volume migration into 2 separate steps:
- `step1`: create snapshot of source volume
- `step2`: clone snapshot + flatten clone + copy clone to destination
  volume

Introduce a parameter which configures, whether `step2` is done while
the VM is off.

Original (non-default) flow:
1. shutoff source VM
2. perform `step1` + `step2` (originally not split)
3. start source VM
4. ...

New default flow:
1. shutoff source VM
2. perform `step1`
3. start source VM
4. perform `step2`
5. ...

Rationale: There is no reason to keep the source VM off once the
snapshot is created. The VM is started after the migration anyway, so
the source and destination volumes will diverge eventually. The new
default approach aims to reduce the time the VM is off.