Allow node shutdowns to change types from "restart" to "remove"

[gwbrown]

As of #71162, node shutdowns can be requested, but cannot be updated by the user after initial creation.

It would be useful if a shutdown could be updated from type RESTART to REMOVAL if the decision is changed partway through shutdown. I don't believe there's anything blocking us from allowing that, beyond updating the API to allow it.

Relates #70338

[elasticmachine]

Pinging @elastic/es-core-infra (Team:Core/Infra)

[sebgl]

I'm wondering about the following use case: long-term, I want to remove a node, so I scheduled a shutdown (removal type).
Short-term: something happened that requires me to restart that same node. This could be anything, but specifically I'm wondering if there are cases where the restart could help the removal (data migration) to go through. For example if that instance suffers from high memory pressure?

Can I turn the removal to a restart? Does it require me to re-request a removal once the restart is done?

From my perspective, those two operations could be decoupled. I want to eventually remove the node (long data migration involved), but right now I need to restart it. The restart intent doesn't change the fact I still want data to be migrated away from the node.