Karatsuba 1.0

[DmytroTym]

This PR introduces a new improved multiplier

[IdoAtlas]

have no comments on code, if it works correctly that's fine.

first, it important that we will have stand-alone benchmark of multiplier for 384 bit field and 256 bit field
each benchmark measures throughput (mults / sec, mod mults / sec) and power (to calculate mod mults / watt). should have those target optimization parameters in front of our eyes at every optimization iteration.

for future improvements:

1. detailed performance breakdown of the mod-mult (how much does each part takes?)
   specific points:

• i encountered a way to get rid of the barrel-shifts, want to see what is the actual impact, if this is a signficant problem let's implement the new method.
• reduce redundency and msb multiplier

2. understanding the karatsuba gains, where do they come from? can we benefit from breaking it another level? can we use ADK for better results?

[DmytroTym]

@IdoAtlas thanks for your comments, I do need to provide benchmarks: standalone multiplier, as well as higher-level primitives. Was going to do it when done with the MSM and interfaces thing, which should be either today or tomorrow

[DmytroTym]

Closed in favour of #289