NUMA instance with shared CPU policy cannot be restarted after upgrade to Victoria

[Pierre Riteau]

Description
===========
NUMA instances without hw:cpu_policy=dedicated set cannot be restarted after upgrading from Ussuri to Victoria. The nova-compute service fails with the following exception:

NotImplementedError: Cannot load 'pcpuset' in the base class

Steps to reproduce
==================
* Deploy Nova using Ussuri
* Create a flavor with hw:numa_nodes='2' but without setting hw:cpu_policy
* Start instance with flavor
* Upgrade Nova to Victoria
* openstack server stop <instance>
* openstack server start <instance>

Expected result
===============
Instance should start.

Actual result
=============
Instance doesn't start.

Environment
===========
python3-nova-22.2.2-1.el8.noarch
openstack-nova-compute-22.2.2-1.el8.noarch

Packaged as Docker images by Kolla.

Logs & Configs
==============
The following trace is logged: https://paste.openstack.org/show/809556/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/openstack/nova/+/810849

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/openstack/nova/+/815923

[Pierre Riteau (priteau)]

This should be resolved by https://review.opendev.org/c/openstack/nova/+/820153

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Pierre Riteau <email address hidden>" on branch: master
Review: https://review.opendev.org/c/openstack/nova/+/810849
Reason: I believe https://review.opendev.org/c/openstack/nova/+/820153 has fixed this issue.