Doc the performance overhead

[littleGnAl]

A rough estimate of the current implementation results in an average CPU overhead of around 1-5%. I’ll need to test on mid-to-low-end devices to obtain more accurate performance metrics.