VMEncryption on CentOS is broken after recent update to the VMEncryption extension

This patch - 27757d457c93b94a54a16519c5e673ffcd797831 which was recently pushed out seems to have broken encrypted CentOS VM's.

If the VM is stopped/started or restarted, then very frequently it will fail to boot. Looking in the boot diagnostics, the boot gets stuck at this point:

> Please enter passphrase for disk Virtual_Disk (osencrypt)

Here is the boot log: [serialconsole.log](https://github.com/Azure/azure-linux-extensions/files/1922402/centos72.9fbf891b-0dc4-4b55-8a7e-13fe55ef52eb.serialconsole.log)

To reproduce:
- Provision a CentOS7.2
`"imageReference": {
"publisher": "OpenLogic",
"offer": "CentOS",
"sku": "7.2n",
"version": "latest"
}`
- Setup a KeyVault which is enabled for disk encryption
- Encrypt the VM like so:
`az vm encryption enable --resource-group --name --aad-client-id --aad-client-secret --disk-encryption-keyvault --key-encryption-key diskencryptionkey --volume-type OS`
- Wait until this command returns a status of VMRestartPending
`az vm encryption show --resource-group --name --query [osDisk] -o tsv`
- Restart the VM, and check the status (it might say to restart again)
- When the VMStatus is Encrypted
- Restart the VM or Stop/Start the VM a few times (Seems to happen more often with Stop/Start), at some point the VM will appear to be hung. Check the boot logs, and you will see that it is stuck on the message "Please enter passphrase for disk Virtual_Disk (osencrypt)"

This is a very serious problem for us (and probably anyone who is running CentOS with encrypted disks on Azure). We stop/start our non-prod environments on a schedule to save costs, this is no longer possible as the restarts are too unreliable. Production environments are now running with a huge risk that any planned or unplanned server restarts will not be recoverable.

There does not be any apparent way to work around this problem.