after client restart, failed task restore leads to cpuset error

While working on another issue, I had a task that failed to restore after restarting the client. When this happened, the cpuset manager started emitting an error once a second into the debug logs:

> 2023-03-23T16:32:15.442-0400 [DEBUG] client.driver_mgr.docker: failed to copy cpuset: driver=docker error="openat2 /sys/fs/cgroup/nomad.slice/d9a1cc4b-f066-b47d-7411-7963a6fa70ac.redis.scope/cpuset.cpus: no such file or directory"

The allocation ID here is the alloc ID of the previously existing allocation that failed to restore. It's replacement allocation has a scope directory as I'd expect.

Ref: https://hashicorp.atlassian.net/browse/NET-12137

Start with the Docker driver and its cpuset manager, then reproduce the failed task-restore flow after a client restart. Trace the handling of the old allocation's missing cpuset.cpus path and compare it with the replacement allocation's scope directory. Done means the restore path no longer emits the repeated cpuset error.

Written by the indexing model from the issue text.