mul and from optimizations

[debris]

• mul is ~30% faster
• from_little_endian / from_big_endian is ~80% faster

before:

test u128_mul      ... bench:     151,032 ns/iter (+/- 32,331)
test u256_from_be  ... bench:          57 ns/iter (+/- 0)
test u256_from_le  ... bench:           51 ns/iter (+/- 1)
test u256_full_mul ... bench:     228,725 ns/iter (+/- 30,183)
test u256_mul      ... bench:     123,390 ns/iter (+/- 32,967)

after:

test u128_mul      ... bench:      98,213 ns/iter (+/- 13,330)
test u256_from_be  ... bench:          13 ns/iter (+/- 0)
test u256_from_le  ... bench:           7 ns/iter (+/- 0)
test u256_full_mul ... bench:     199,849 ns/iter (+/- 19,195)
test u256_mul      ... bench:     105,883 ns/iter (+/- 14,433)

pleasereview @NikVolf / @Vurich

[NikVolf]

Wow, this is awesome, thanks!

[NikVolf]

I've just realized we don't have a test that verifies that from_* functions actually do work in the right order

care to merge pr #24 into your one to actually check this?

[debris]

sure, done