Feat: 4-dimensional fake GLV

[yelhousni]

Description

https://ethresear.ch/t/fake-glv-you-dont-need-an-efficient-endomorphism-to-implement-glv-like-scalar-multiplication-in-snark-circuits/20394/7

Type of change

• [x] New feature (non-breaking change which adds functionality)

How has this been tested?

Correctness tests are implemented in emulated/sw_emulated/point_test.go and native/sw_bls12377/g1_test.go.

How has this been benchmarked?

All benchmarks are in plonkish constraints (scs).

• Emulated scalar multiplication in a BN254-PLONK:

scalar mul	old (scs)	new (scs)
Secp256k1	385,461	282,223
BN254	381,467	279,262
BW6-761	1,367,067	1,010,785
BLS12-381	539,973	390,294

• Native scalar multiplication on BLS12-377 in a BW6-761-PLONK:

scalar mul	old (scs)	new (scs)
BLS12-377	3,587	3,567

• Ethereum precompiles circuits:

precompiles	old (scs)	new (scs)
ECMUL (BN254)	384,782	280,123

• PLONK recursion:

Recursion	old (scs)	new (scs)
BW6-761 in BN254 (emulated)	28,206,642	27,137,821

Checklist:

• [x] I have performed a self-review of my code
• [x] I have commented my code, particularly in hard-to-understand areas
• [x] I have made corresponding changes to the documentation
• [x] I have added tests that prove my fix is effective or that my feature works
• [x] I did not modify files generated from templates
• [x] golangci-lint does not output errors locally
• [x] New and existing unit tests pass locally with my changes
• [x] Any dependent changes have been merged and published in downstream modules

[yelhousni]

I'm not fully sure about the possible overflow in BLS12-377, but I think with this approach we have a slight problem. Imo it looks good for the emulated 4-dim, but not for 2-chain.

Applied the suggestions and resolved conversations. For BLS12-377, we only use the GLV+Fake-GLV for testing purposes as it is currently more expensive constraint-wise compared to classical GLV. If we do the scalars decomposition check in non-native then definitely way more expensive.

[ivokub]

Thanks for all the changes! All good from my side now. I'll postpone the general interfaces for different curves for a separate PR (https://github.com/Consensys/gnark/issues/1303)