Separate nodeSelector logic for scheduler+webserver vs ephemeral resources

[aimran-adroll]

Description
I would like to separate out the current nodeSelector logic for critical (imo) pods such as the scheduler+webserver vs worker pods.

Use case / motivation

If we look at the current implementation,

airflow/chart/values.yaml

Lines 37 to 40 in ba36f87

	# Select certain nodes for airflow pods.
	nodeSelector: {}
	affinity: {}
	tolerations: []

we have only one place to specify the logic above

Subsequently, the above template is applied as is across all the resources (e.g scheduler, webserver, worker, ...)

Typically, I work with an autoscaling group that consists of few stable (ie on-demand) instances along with the more ephemeral spot (preemptible) instances.

During deployment, I would like to assign the critical (TBD but at least scheduler and webserver) pods to the on-demand instances.

Related Issues

grepping for nodeSelector did not turn up anything here or the astronomer chart repo.

[kaxil]

Closed by #13508