[roll] Roll fuchsia [audio_core] Add an output pipeline benchmark

This CL adds a new command line tool that constructs an OutputPipline
from a given audio_core_config.json file. It then benchmarks
OutputPipeline::ReadLock against a set of input streams. This program
will be used to verify that out-of-process effects (fxbug.dev/80067) do
not slow down our production audio pipelines by a significant amount.

The program ships with a default set of configurations:

  empty (no inputs)
  M/VC (one MEDIA input, constant volume)
  C/VC (one COMMUNICATION input, constant volume)
  U/VC (one ULTRASOUND input, constant volume)
  MCU/VC (one of each of the above inputs, constant volume)
  MCU/VS (one of each of the above inputs, stepped volume)
  MCU/VR (one of each of the above inputs, ramped volume)

The inputs are sine waves. For audible inputs, we use a 1kHz wave
sampled at 48kHz with 1 channel. For ultrasonic inputs, we use a 32kHz
wave sampled at 96kHz with 1 channel. These inputs are currently not
configurable. However, the set of usages is configurable. For example,
`--scenarios=MMI,U` benchmarks a scenario with 2 MEDIA inputs and 1
INTERRUPTION input, plus a second scenario with 1 ULTRASOUND input.

The tool is designed to run against an arbitrary audio_core_config.json.
To support this, we define the BUILD rules within a template which must
be instantiated for each audio_core_config.json. For now, this template
is instantiated just once, with an empty pipeline. A future CL will
instantiate this template with our production pipelines.

Below is the output from the tool when run on an empty output pipeline
on a sherlock device.

$ fx shell run fuchsia-pkg://fuchsia.com/audio-pipeline-benchmark-empty-pipeline#meta/audio-pipeline-benchmark-empty-pipeline.cmx
Audio output pipeline profiling tool

    Elapsed time in microseconds for a single 10.00ms mix job
    for mixer configuration X/VV where X is a list of input
    streams, each of which has one of the following usages:

        B: BACKGROUND
        M: MEDIA
        I: INTERRUPTION
        S: SYSTEM_AGENT
        C: COMMUNICATION
        U: ULTRASOUND

    and VV is a volume setting:

        VM: muted volume
        VC: constant volume
        VS: discrete volume change just before each mix job ("stepped")
        VR: ramped volume change just before each mix job

Config        Mean          Median            Best           Worst      Iterations
empty       41.583          42.500          15.333          49.083              20
M/VC       346.366         343.208         329.417         396.250              20
C/VC       341.612         337.083         328.125         402.042              20
U/VC        84.772          82.208          72.042         105.541              20
MCU/VM     113.502         108.666          90.542         210.959              20
WARNING: 20 of 20 runs produced no output
MCU/VC     643.995         635.854         625.792         782.125              20
MCU/VS     620.089         614.312         596.167         712.708              20
MCU/VR     991.877        1007.062         813.375        1049.208              20

Original-Bug: 80067
Original-Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/574060
Original-Revision: 57c950c09b70bdebed6838259e3d77edf3d564bb
GitOrigin-RevId: 6d6b00a3acaddaa0f527909f10064aa30f86a71b
Change-Id: I646959240961460f23903be2b9410ec01ccf94e1