Test performance of obfuscated chacha20 and compare with expected

[nobugs-hare]

Expected: OBF3 - expected increase of 30x, OBF4 - of 100x.

Measured times (on chacha_cipher(), Random Tests (Release))

• Everywhere below: no obf/OBF3/OBF4 Windows/MSVC/x64 (box1): 250ns/1200-5000ns/4000-9000ns = 1x/5x-25x/20x-35x Windows/MSVC/x86 (box1): 280ns/1800-5000ns/4000-10000ns = 1x/6x-20x/15x-35x Mac OS/Clang/x64 (box 2): 20ns/500-1000ns/1200-3000ns = 1x/25x-50x/60x-150x Mac OS/Clang/x86 (box 2): 25ns/500-1100ns/1000-3500ns = 1x/20x-45x/40x-140x Linux/GCC/x64 (box 3): 200ns/1100-2000ns/2300-6500ns = 1x/5x-10x/12x-32x Linux/GCC/x86 (box 3): 300ns/800-2000ns/4000ns = 1x/2.5x-7x/15x

With "an order of magnitude" precision, these are "good enough" for our purposes. Further work to make ad-hoc estimates more precise is possible, but for the time being this will certainly do.