Add startup mode to benchmarks

[pjBooms]

Introduced a new startup benchmark mode that measures application startup performance, from process start to the first frame and beyond.

Fixes CMP-9700

Release Notes

N/A

[ASalavei]

LGTM. Couple of notes:

1. The measurement looks accurate, however it must be done with release mode and disabled debugger.
2. Fist launch much slower then the following launches (cold vs hot launch) - this must be considered during launch tests.
3. iOS device behaves differently with attached/detached cable. To ensure that the numbers measured correctly, I would recommend to do several launches of the app not from XCode, but with the standard way - by tapping an icon in both: cold and hot start.
4. Would be great to collect these metrics on a regular bases from one of our apps (KotlinConf?)

[pjBooms]

LGTM. Couple of notes:

1. The measurement looks accurate, however it must be done with release mode and disabled debugger.
2. Fist launch much slower then the following launches (cold vs hot launch) - this must be considered during launch tests.
3. iOS device behaves differently with attached/detached cable. To ensure that the numbers measured correctly, I would recommend to do several launches of the app not from XCode, but with the standard way - by tapping an icon in both: cold and hot start.

Sure. The purpose of the benchmarks is to be installed in the performance infrastructure to catch regressions/improvements. We will take into account these notes (/cc @PavelPunegov)

4. Would be great to collect these metrics on a regular bases from one of our apps (KotlinConf?)

I don't think that real apps are ok to add to the benchmarking system, though I agree that we could measure similar way startup of them separately (as an additional task).