[uapp][trace-benchmark] Commonize benchmark handling, improve fidelity

Treat "notrace" and "disabled" tests the same as "normal" tests.

Plus rerun benchmarks multiple times and report min,max,average
to cope with excessive randomness in results.

Tested: trace-benchmark

Change-Id: I0fadfd3845d2038fb0cfc93a9b9eff93d6c373e9

[tracing] Add support for circular and streaming modes

Tested:
/boot/test/sys/trace-test
trace-benchmark
Further testing done on the garnet side.

Change-Id: I8ec21ba0adc57d090bbfe4be6d56fc695aa645d4

[tracing] Add beginnings of trace buffering mode support

This patch just adds the basic infrastructure.
Support for circular and streaming modes comes later.

Tested:
/system/test/sys/trace-test
trace record --buffering-mode=oneshot \
--categories=stress:example,stress:something \
/system/bin/trace-stress --count=5000
trace record --buffering-mode=circular \
--categories=stress:example,stress:something \
/system/bin/trace-stress --count=5000
[trace will be empty, but should run]
trace record --buffering-mode=streaming \
--categories=stress:example,stress:something \
/system/bin/trace-stress --count=5000
[trace will be empty, but should run]

Change-Id: I1fc28668dee652c88df9d5e669c8b4ac978604bc

[async] Change async_t -> async_dispatcher_t

This removes the compatibility wrappers once all the higher layers have
been migrated.

Bug:ZX-2337
Test:CQ
Change-Id: If8b59c157bbfadc63b426e679bce641ee7a7abd0

[uapp][trace-benchmark] Reorganize to prepare for streaming mode

- the buffer was filling which invalidates results
- run each benchmark with a fresh buffer
- prepare to support for running under different modes
(oneshot, circular, streaming)
- reduce iteration count so buffer doesn't fill
1,000,000 -> 100,000 shouldn't affect reported results

Tested: trace-benchmark

Change-Id: I171ca856fd488d5430c9e8fba2a4f85f6107a66f