split samples into multiple versions: simple and optimized

As part of porting our samples to ARM, we're seeing them become more
complex with ifdefs around the arch-specific optimized parts. One
proposal is to split them up: have a simple version of the sample, and
then optimized versions that are perhaps split into _x86 and _arm files.

bbbuf_simple = clean call
bbbuf_opt1 = inc full ptr, check for end
bbbuf_opt_x86 = if opt approach different than arm