Bring back metrics and trace

[gmarz]

Two features from 1.x that have been removed in master due to refactoring are MetricsEnabled() and TraceEnabled(). We should add them back, and I think trace can leverage the new AuditTrail bits?

[Mpdreamz]

I'm fine with leaving Trace out, i rather introduce LibLog and log .DebugInformation on INFO.

Our audit trail already gives timings for each individual step in the pipeline but not per step what the time to first byte / attempted serialization/deserialization time was.

This would be awesome to bring back through metrics, however if we can do it always in a low allocation fashion that'd be awesome++.

[maeserichar]

So in case we want to log the API requests, what should be the way in the 2.x version?

[Mpdreamz]

Specifiy a handler on ConnectionSettings for OnRequestCompleted()

[maeserichar]

Great, thanks!

[Mpdreamz]

Closing this as the path forward is DiagnosticSource for tracing and EventCounter for metrics. Both of which have landed in the client and we'll look to augment, stabalize and document for the 8.0.0 release.

cc @russcam @codebrain