ecdsa library optimization

i notic that starkware is not using the secp256k1 curve.

however, i compared the ecdsa sign to bitcoin library secp256k1, it is much slower. libsecp256k1 only takes 85 micro per iteration.

after some digging, the current implementation does not adopt the jacobian coordinates as in libsecp256k1.

any plan to optimize that?