The gRPC client for CSI connected to the socket during initialization
so that we can ensure the gRPC service interfaces have been
initialized before use. But this design makes it challenging to
correctly retry and timeout the initial connection in the plugin
supervisor and then reuse that same client for fingerprinting.

Lazily connect to the socket on the first RPC call, allowing the
caller's context to be used to control retries and timeouts.

This also gives the CSI plugin manager instance the ability to retry
its initial connection setup.

The plugin supervisor hook creates a new gRPC client for every probe
and then throws it away. Instead, reuse the client as we do for the
plugin manager.

The plugin supervisor registers the plugin in the `Poststart` hook, so
the task itself should be running. If the plugin can't communicate
with us after 30s, exit and mark the task as unhealthy so that it can
be restarted.