yuv2buf for processing ImageReader output

[gordinmitya]

Sorry, I unintentionally closed #336 and had to close #404 because of google-cla bot. So here is updated version of them.

Part discussion of this code is in issue #277.
Documentation about approach, unit-tests, performance benchmarks + demo application are here https://github.com/gordinmitya/yuv2buf. This code is tested in several apps in production with both: Camera2 output and MediaCodec outputs.

I compared performance of previous and new code on my Mi 9t pro (Snapdragon 855): old 3.5-4.0ms, new 2.6-3.0ms.
Modified CameraActivity.kt was used for comparison.

Will be happy to answer questions, if any!

[ggfan]

cool. a separate note, if you have not filled the CameraX 1.0 survey yet, may you fill it here: https://forms.gle/DarkBYwWFmDzdtLr8? It would be a good one to Camera(X) team: make sure to state your suggestions! Thanks!

[ggfan]

Confirmed for the performance gain in Pixel3 devices with images 640x480: about 30% improvement. Performance is import. A couple of things to do, I think:

• Add some more detailed comment for YuvToRgbConvert() in your version or keep the original warning there (feel free to change but keep the warning might be good)
• Convert the file to Kotlin
• CameraX dependency: need to decide, punt it for now.

could you help for them? thx

[ggfan]

@gordinmitya merging this one now, then I will get help internally to convert it to kotlin then plug it into samples (right now it is in the lib module, not used). thank you for the PR.