improve: avoid the inversions in iso_map_secp256k1

[duguorong009]

Description

Improve the iso_map_secp256k1 func by avoiding the scalar inversions at all.

Related issues

Changes

• use the jacobian coordinates instead of affine one in the func
• update the iso_map logic by using jacobian coordinates

[davidnevadoc]

LGTM! 👍

Just a couple of suggestions:

1. Let's add an RFC reference for this modification https://www.ietf.org/archive/id/draft-irtf-cfrg-hash-to-curve-10.html#section-appendix.e

2. I think we no longer need to follow the naming convention for the constants K since we are following a different algorithm. I would remove the first [Fp::ZERO, 4] and add a comment specifying that K[0][..] are x_num constants, K[1][..] are x_den constants, ...

These suggestions are optinal, feel free to merge!

[CPerezz]

Can we try to merge this prior to the release in #125 ??

[duguorong009]

LGTM! 👍

Just a couple of suggestions:

1. Let's add an RFC reference for this modification https://www.ietf.org/archive/id/draft-irtf-cfrg-hash-to-curve-10.html#section-appendix.e
2. I think we no longer need to follow the naming convention for the constants K since we are following a different algorithm. I would remove the first [Fp::ZERO, 4] and add a comment specifying that K[0][..] are x_num constants, K[1][..] are x_den constants, ...

These suggestions are optinal, feel free to merge!

Thanks for quick review! @davidnevadoc

1. We already have the RFC reference in here.
2. Even though we use the algorithm to avoid inversion, the constants meanings are not different from original one.
   I think it is okay to keep the current K as is.

Hence, I would go for merge!

[duguorong009]

Can we try to merge this prior to the release in #125 ??

Yes, I will merge the PR right now since @davidnevadoc gave approval.