Misc rdkafka orderer updates

[GaryWilber]

• Remove batching logic from the producer - we're all always setting MaxBatchSize to 1 to avoid kafka message size issues. The producer can be simplified by removing that logic
• Add a reconnectOnNonFatalErrors option to the producer - the librdkafka producer is smart and automatically retries message producing most of the time. It's not necessary to close and re-open it on every error it gets
• Tweak how the producer tracks the connecting / connected state
• Fix consumer errors during close / server shutdown
• Add support for disabling automatic topic creation - this logic must be disabled when targeting Azure Event Hubs
• Add support for rdkafka log events