[pyproto] error handling and built-in retry strategy

[yunhaoling]

Proposal:

• error condition code driven retry mechanism
• strategy for different layers failures:
  • Transport layer: wrapped in AMQPConnection, reestablish every layer and retry
  • AMQP connection: connection close, reestablish every layer and retry
  • AMQP session: session end, reestablish session/link/client and retry
  • AMQP link: link detach, reestablish the link/client and retry
  • AMQP client level: send fail/ receive fail, no need to reestablish, retry the operation
  • Other errors: no close, no retry, just raise(?)
• errors/retry are handled at the client layer
• features need to be retried:
  • send
  • receive
  • mgmt request
• custom error policy for EH and SB, client takes:
  • 1) general retry configs, retry_total/max_backoff/etc. (some of azure-core configs)
  • 2) a map in which key is error condition code, value is the backoff time
  • 3) a list of non retryable error code