[Metrics] Improve ES Output metrics

[lucabelluccini]

The common problem in Logstash performance cases is to determine if there is slowness in Logstash itself, the inputs or if there’s pushback / backpressure from the outputs.
I think it would be valuable to enhance the Elasticsearch output, adding some histograms/stats about:

• Number of concurrent requests / in flight to Elasticsearch
• Average/Distribution of RTT of indexing request to Elasticsearch
• Occurrence of errors (retriable, not retriable), possibly with status codes (429, 503, 502…)
• Distribution of bulk requests sizes (both in terms of bytes per bulk, in terms of number of documents per bulk)
• Whenever possible, distribution of different data streams or index names involved per bulk
• Other useful metrics