Improve error logging via events when attach fails on a node

[gnufied]

I have seen many error reports where volume fails to attach/mount on a node with following generic error:

"Error syncing pod, skipping: timeout expired waiting for volumes to attach/mount for pod "mongodb-2-z5dbt"/"foobar". list of unattached/unmounted volumes=[mongodb-data]"

and further debugging often requires access to both controller and kubelet logs, but those can be tricky to acquire. The user who is running into this bug may not have access to controller logs.

We need to make it simpler to identify all instances of failures and why it is happening. There have been some cases such #42006 which improves it, but we should try and cover all branches of code to make sure - an error is simply not dropped on the floor.

cc @kubernetes/sig-storage-bugs

[gnufied]

I am going to close this bug for now. We do have necessary logging that helps debugging this problem. There was unrelated error that prevented log events to be showing up sometimes.