Add configurable nodeSelectors to Fluid core workloads

[gdagil]

Summary

• Add default kubernetes.io/os: linux nodeSelectors for all core Fluid workloads (controllers, dataset, runtime controllers, CSI daemonset, webhook, CRD upgrade job) in values.
• Wire nodeSelector blocks into every relevant pod template so users can override or extend scheduling constraints.
• Keep existing tolerations and affinity behavior unchanged.

Changes

• values.yaml: new nodeSelector defaults for dataset, CSI, webhook, fluidapp, CRD upgrader, and each runtime controller (alluxio, jindo, goosefs, juicefs, thin, efc, vineyard).
• Templates: inject nodeSelector into controller Deployments, CSI DaemonSet, webhook Deployment, and CRD upgrade Job.

Testing

• helm template charts/fluid (manual) to confirm each rendered pod includes the nodeSelector stanza.

Notes for reviewers

• Defaults are minimal and match Kubernetes’ standard label; users can override per component via values.yaml.
• No behavior change for clusters already scheduling on Linux; adds explicit scheduling hook for constrained clusters.

Checklist

• Changes limited to Helm values/templates for scheduling.
• Tested with helm template.
• Commit includes Signed-off-by.

Branch

feature/add-node-selector

How to override (example)

dataset:
  nodeSelector:
    kubernetes.io/os: linux
    my.custom/zone: az1