fix: override systemd unit

[bcho]

This pull request refactors how systemd unit and drop-in files are managed across several components, introducing idempotent "ensure" methods that only write files when necessary and centralizing logic for starting and reloading units. The changes improve reliability and reduce unnecessary restarts or reloads by tracking when files have actually changed.

Systemd file management improvements:

• Replaces previous WriteUnitFile and WriteDropInFile methods with new EnsureUnitFile and EnsureDropInFile methods in the systemd.Manager interface, which only write files if their content has changed or if the file does not exist. These methods now return a boolean indicating whether the file was updated. (pkg/systemd/types.go, pkg/systemd/dbus.go) [1] [2]
• Updates all usages in kubelet, kubeadm, and cri components to use the new "ensure" methods, removing manual checks for file existence and content. (components/kubelet/v20260301/kubelet_service.go, components/kubeadm/v20260301/join.go, components/cri/v20260301/service.go) [1] [2] [3] [4]

Unit lifecycle and reload logic centralization:

• Introduces a new EnsureUnitRunning helper function that encapsulates logic for reloading the systemd daemon and restarting or starting units only when necessary, based on whether unit files or drop-ins were updated. (pkg/systemd/ensure.go)
• Refactors component logic to use this new helper, simplifying code and ensuring consistent behavior across services. (components/kubelet/v20260301/kubelet_service.go, components/cri/v20260301/service.go) [1] [2]

Code cleanup:

• Removes unused error handling and file existence checks, as these are now handled by the new "ensure" methods. (components/kubelet/v20260301/kubelet_service.go, pkg/systemd/dbus.go) [1] [2]

These changes make systemd unit management more robust and reduce unnecessary operations, improving both reliability and maintainability.