fp12_tower.c: add blst_bendian_from_fp12.

[dot-asm]

See #101 for details.

[mratsim]

I still think it's better to use the canonical representation instead of assuming Fp2 -> Fp6 -> Fp12 towering.

I found a paper mentioning it as well:

chapter 6.4

• Multiplication and Squaring on Pairing-Friendly Fields
  Augusto Jun Devegili and Colm Ó hÉigeartaigh and Michael Scott and Ricardo Dahab, 2006
  https://eprint.iacr.org/2006/471

[dot-asm]

I still think it's better to use the canonical representation instead of assuming Fp2 -> Fp6 -> Fp12 towering.

And that's what I aimed for. What did I miss? Note that loops are in "inverse" order. I mean for non-canonical representation the outer loop would have 2 iterations...

[MariusVanDerWijden]

LGTM

[MariusVanDerWijden]

Friendly bump.
We would love to add BLST to our differential fuzzer that runs on oss-fuzz and the only thing missing is this PR

[dot-asm]

Committed. Thanks for the feedback! And the bump ;-)