Improve source mount recovery reliability

[vladem]

Issue #, if available: N/A

Description of changes:

1. Add a test for mount recovery. Note that the workload continuously accesses (cat /mnt/volume1/file.txt) the dead mount, which might be a rare scenario.
2. Update CheckMountpoint to detect dead mount when the stat(target) is cached in kernel
3. Update behaviour of aws-s3-csi-mounter when waiting for mount options: instead of exiting and restarting (which might result in failed mount attempts, when backoff becomes large), it continues running until an .exit file was placed by csi-node

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.