panic when getting worker parallelism for non CN nodes

[xxchan]

          This will panic in some places:

• diagnose in dashboard
• show cluster

Can you examine the usages of worker.parallelism()? (Or you may change list_all_nodes() calls to list_cn())

Originally posted by @xxchan in #19153 (comment)

[shanicky]

The previous design only considered parallelism for compute nodes, so an assertion was made. Is it possible to determine the type of the server before making a call?

[xxchan]

e.g., ListAllNodesRequest contains a worker_type parameter