ETW tracing improvements

[sebastianburckhardt]

Several minor improvements to simplify analytics:

• do not trace batchworker work cycles when the batch is empty
• use 'Details' instead of 'Reason' for checkpoint events
• include partition id in EventHubs traces
• use LatencyMs instead of Latency for batch worker tracing
• Do not emit warnings for ReceiverDisconnectedExceptions observed inside EventProcessorHost