Move calculation of disk_available_least to resource tracker

Currently in libvirt[1] and xenapi[2] driver, the 'get_available_resource()'
method will report 'disk_available_least' in return value.
[1]https://github.com/openstack/nova/blob/13.0.0.0b3/nova/virt/libvirt/driver.py#L5416
[2]https://github.com/openstack/nova/blob/13.0.0.0b3/nova/virt/xenapi/driver.py#L458

However, libvirt and xenapi driver only take the instances known by hypervisor
into account but ignore the instances being migrate to/from the host.

To solve this problem, We need to move the calculation of
'ComputeNode.disk_vailable_least' to resource tracker, because only
the resource tracker holding the global resource lock knows the current
usage of resources since it can take both existing instances and
ongoing migrations into account.