gpu: plugin exit gracefully on host without particular HW device.

[SidneyAn]

intel-gpu-plugin will crash on host without Intel GPU (i915 driver) like following.
controller-1:~$ kubectl logs intel-gpu-plugin-8sqch -n kube-system
GPU device plugin started
Device scan failed: open /sys/class/drm: no such file or directory
Can't read sysfs folder
main.(*devicePlugin).scan
/go/src/github.com/intel/intel-device-plugins-for kubernetes/cmd/gpu_plugin/gpu_plugin.go:83
main.(*devicePlugin).Scan
/go/src/github.com/intel/intel-device-plugins-for-kubernetes/cmd/gpu_plugin/gpu_plugin.go:69
github.com/intel/intel-device-plugins-for-kubernetes/pkg/deviceplugin.(*Manager).Run.func1
/go/src/github.com/intel/intel-device-plugins-for-kubernetes/pkg/deviceplugin/manager.go:96
runtime.goexit
/usr/local/go/src/runtime/asm_amd64.s:1333

Shall we catch the exception and print some helpful message like "Failed to scan device. If this is a GPU node, you can check the prerequisites at: https://....." instead?

[grahamwhaley]

Looks like we should at least fail more gracefully.
OOI @SidneyAn - how did you deploy the plugin to the node(s) - that may make some difference as well to how it is handled, as if via a daemonset, I think if not handled carefully then the daemonset can get stuck in a retry loop.