Provide simpler Prometheus metrics in worker clients

[aklenik]

Context

The worker clients provide some aggregated metrics related to the TXs since the last push/update. This is not the Prometheus way, as the Prometheus queries usually specify their own aggregation window.

Expected Behavior

Specify only the following metrics:

• caliper_tx_submitted: counter, incremented when submitCallback is called, with the client id as a label.
• caliper_tx_finished: counter, incremented when the callback returns the result array, with the client ID and final status (success/failed) as labels
• caliper_tx_e2e_latency: histogram, updated when caliper_tx_finished is updated (histograms are cheap on the client-side, unlike summaries)

A number of metrics may be derived from the above using Prometheus queries, like goodput, failure rate, total TPS (successful and failed), sending TPS and pending transactions. With little to no performance cost for the worker clients.

Actual Behavior

Metrics are locally aggregated, then pushed.

Your Environment

• Version used: 0.2.0