[trace_event] tracking issue

This is a broad tracking issue for graduating Trace Events from experimental status. Start by reading the initial Trace PR and the linked issues #81, #82, and #83, then choose one unchecked item with a defined scope. Done is completing that selected task and updating its checklist status.

This issue is to track ongoing work to refine Trace Events in Node and graduate it from experimental state.

It builds on and replaces

• https://github.com/nodejs/diagnostics/issues/30
• https://github.com/nodejs/diagnostics/issues/53

Initial Trace PR including discussion: node#9304

todo:

• configurable output destinations
  • support targeting lttng, etw, dtrace, etc...
• add a listener system (#81)
• C/C++ API (#83)
• JavaScript API (node-eps#48)
  • needs to be as fast as possible, probably a VM intrinsic.
• add trace points to nodejs/node (#82)
• move existing dtrace/lttng trace writes to trace_event
• add trace points: async hooks (PR: https://github.com/nodejs/node/pull/15538)
• relationship to upstream trace_event_common.h (see https://github.com/nodejs/node/pull/10628)
• recommendations for naming categories for ecosystem
• need approach on how to collect "context"
  • e.g., for capturing http connection info, currently connection object is passed through DTrace APIs, and some handler/provider, and then the handler/provider will navigate connection object to pull out interesting data to trace. This approach is brittle (dependent on structure of connection object), and puts burden on provider (ie., won't easily allow using different event tracing systems like lttng, dtrace,...).
• standardize on output format for trace event data
• command-line options to support tracing features.