'vmdk-opsd stop' needs to drain operations in flight

[msterin]

Currently vmdk-opsd stop just stops (exits) the vmdk_ops.py service. If there were operations in fligth it could kill them in the middle of execution. That can potentially create inconsistencies in VM attachement, KV files or auth-db.

To prevent this, we need to drain the ops in flight before stopping.
Something along this line
[ ] On 'stop' requests, add a barrier so executeRequest does fails all new incoming requests with "The server is stopping, operation declined".
[ ] Add count of ops in flight, and when barrier is installed makes sure the count drops to zero before proceeding
[ ] Add a way to force-kill the service in case something is stuck (e.g. a 10TB format is running). This will behave as today's "stop" but issue additional message about potential danger. I suggest vmdk-opsd kill.
[ ] Add a message for "status" that the "service is being stopped, waiting for completion of X operations in flight "

//CC @pdhamdhere

[shaominchen]

[ ] Add a message for "status" that the "service is being stopped, waiting for completion of X operations in flight "

As per offline discussion with @msterin, we think it makes more sense to print the message for "stop" itself instead of doing that for "status".