Size JVM pools based on node.processors

[henningandersen]

Description

In #89662 we changed the node.processors setting to be a double, enabling precise representation of the environment (often docker container) that ES is running in.

However, we still require configuration of the JVM thread pools, preferably through -XX:ActiveProcessorCount. In order to simplify it would be good to instead derive this from the node.processors setting such that the only setting we need about the environment is that setting. This is slightly complicated because it is in the startup process, but seems doable to parse out the setting from the yaml.

Relates #65905 and #89662

[elasticsearchmachine]

Pinging @elastic/es-delivery (Team:Delivery)

[mark-vieira]

This is slightly complicated because it is in the startup process, but seems doable to parse out the setting from the yaml.

The complex bit is that you can pass settings via multiple ways, not just elasticsearch.yaml. Right now our auto-heap is somewhat naive and assumes that node roles will only exist in the YAML config and not via environment variables, CLI args, etc.

[elasticsearchmachine]

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

[rjernst]

This was addressed with #101846