When attempting to cleanup the existing connections to the VSphere
cluster, we get the driver from the state shared between all steps in
order to run the commands necessary for that.

This can however fail, if the step couldn't succeed for any reason.

Then, if the connection fails and the driver isn't set in the state,
when running Cleanup on the connect step, we get the driver from state
and cast it to a driver.Driver.
This will make the plugin crash, and so we change how we proceed here,
by getting the driver with a `GetOk`, exiting if it failed to be fetched
from state, then casting the result if it was created in state.

This should prevent crashes like those.

Co-authored-by: Ryan Johnson <johnsonryan@vmware.com>