perf: add optional support for keccak-asm

[DaniPopes]

Motivation

Closes #414
See https://github.com/DaniPopes/keccak-asm, https://github.com/DaniPopes/bench-keccak256

Solution

Add "asm-keccak" feature which overrides the tiny-keccak implementation with keccak-asm

Benches

https://github.com/DaniPopes/bench-keccak256

Local:

RUSTFLAGS=""
2.60-2.75 µs -> 2.10-2.20 µs

primitives/keccak256/32 time:   [2.1212 µs 2.1285 µs 2.1373 µs]                                     
                        change: [-21.731% -21.336% -20.849%] (p = 0.00 < 0.05)
                        Performance has improved.

RUSTFLAGS="-Ctarget-cpu=native" (AVX512VL)
2.08-2.20 µs -> 2.04-2.10 µs

primitives/keccak256/32 time:   [2.0734 µs 2.0747 µs 2.0761 µs]                                     
                        change: [-5.1237% -5.0213% -4.8902%] (p = 0.00 < 0.05)
                        Performance has improved.

PR Checklist

• Added Tests
• Added Documentation
• Breaking changes