EAI-8205 Give every node a local-path storage path

[pre]

Related:

• EAI-8203 Support Ubuntu 26.04 #306

Bloom wrote the local-path provisioner config with a nodePathMap that names
the first node of the cluster:

{"nodePathMap":[{"node":"ubuntu24-1","paths":["/mnt/disk0"]}]}

A node that joins later got no path. When the scheduler put a pod with a volume
on a join node, the provisioner refused it:

failed to provision volume with StorageClass "direct": config doesn't contain
node ubuntu24-2, and no DEFAULT_PATH_FOR_NON_LISTED_NODES available

The volume stayed Pending, and the pod with it. On a two-node cluster this
stopped the cluster-forge deployment at the OpenBao volume, after every Bloom
task passed.

The config manifest already defaults to DEFAULT_PATH_FOR_NON_LISTED_NODES,
which applies the paths to every node. The task that read the first node name
overrode that default, so this change removes the task and the variable. A node
that joins later needs no change now.

Bloom gives each node the same /mnt/diskN paths from its CLUSTER_DISKS, so
one path list fits the whole cluster.

Test

Found and fixed while testing #306 on two two-node clusters, one Ubuntu 24.04
and one Ubuntu 26.04. Both showed the fault, so it is not specific to an Ubuntu
release.

After the change, on both clusters:

• The config holds
  {"nodePathMap":[{"node":"DEFAULT_PATH_FOR_NON_LISTED_NODES","paths":["/mnt/disk0"]}]}.
• Bloom's own storage test volume is created on the join node:
  Volume pvc-... has been created on ubuntu24-2:/mnt/disk0/....
• cluster-forge v2.2.2 deploys to the end, and every volume binds.

Jira: https://amd.atlassian.net/browse/EAI-8205

[lauri-amd]

There's an assumption that the all the nodes have the same amount of disks available at the same paths?
If the initial has 2 disks the CM is going to be /mnt/disk0, /mnt/disk1 regardless of how many the other nodes have. The other one might have just single disk at /mnt/disk0 .
I think this is still a better than the current state, ✅