Cleanups and optimizations

See commit messages for more info.

I did some tests on my OPI3 board using the same scenario (expand/collapse status bar for a 5 times). Inferno results:

rework+cache.html (file size 2.1MB)

• drm_hwcomposer: Implement BI and FB caching
• drm_hwcomposer: Add unique handle ID getter into buffergetter class

rework-only.html (File size 2.7MB)

• drm_hwcomposer: Reorganize struct DrmHwcLayer
• drm_hwcomposer: Set return type to std::optional for BufferInfoGetters
• drm_hwcomposer: Move include/drmhwcgralloc.h to bufferinfo/BufferInfo.h
• drm_hwcomposer: Use atrace for framebuffer's import/remove events

baseline.html (File size 3.8MB)

• (main) drm_hwcomposer: Add non-blocking commit support

Simpleperf result file sizes itself tells a lot about optimizations made.