if use gpu-operator install nvidia driver, when node restart, driver-plugin start CrashLoopBackOff

[lengrongfu]

1. when the node reboots after, gpu-operator can reinstall NVIDIA driver;
2. but the same time device-plugin pod starts, but this driver does not install completely. so pod in CrashLoopBackOff.

we can add a initContainers to check nvidia driver to resolve this problme.

[lengrongfu]

/assign

[github-actions]

This issue has been automatically marked as stale because it has not
had recent activity. It will be closed if no further activity occurs.

[lengrongfu]

User Case：

• User use gpu-operator install driver and toolkit, then node reboot, hami-device-plugin can restart pod success.
• User use gpu-operator install driver and toolkit, then delete driver pod, retry install new driver, hami-device-plugin can restart pod success.
• User use manual install driver and toolkit, then node reboot, hami-device-plugin can restart pod success.