Issue 7036: node selection for data mover backup

[Lyndon-Li]

Fix issue #7036. Add the implementation of node selection for data mover backups

[codecov]

Codecov Report

Attention: Patch coverage is 84.68468% with 17 lines in your changes are missing coverage. Please review.

Project coverage is 61.74%. Comparing base (1034d6a) to head (dccde10).
Report is 64 commits behind head on main.

Files	Patch %	Lines
pkg/cmd/cli/nodeagent/server.go	58.82%	7 Missing ⚠️
pkg/exposer/csi_snapshot.go	92.85%	3 Missing and 1 partial ⚠️
pkg/exposer/generic_restore.go	78.94%	3 Missing and 1 partial ⚠️
pkg/controller/data_download_controller.go	80.00%	0 Missing and 1 partial ⚠️
pkg/controller/data_upload_controller.go	87.50%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7437      +/-   ##
==========================================
+ Coverage   61.73%   61.74%   +0.01%     
==========================================
  Files         262      263       +1     
  Lines       28429    28780     +351     
==========================================
+ Hits        17550    17770     +220     
- Misses       9646     9763     +117     
- Partials     1233     1247      +14     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ywk253100]

Why we cannot use the corev1.Affinity directly rather than defining our own type?

[Lyndon-Li]

I think there are two questions:

• Why do we wrap metav1.LabelSelector/corev1.Affinity instead of using it directly? -- We may need to add some other controllers that metav1.LabelSelector/corev1.Affinity cannot provide in future, so we add a simple wrap here. Whenever we want to change, we just append new fields so as to avoid break changes.
• Why don't we use corev1.Affinity directly? -- corev1.Affinity is more complex than we want to provide to users, e.g., we only use RequiredDuringSchedulingIgnoredDuringExecution, so we need to limit users input