Speed up proof generation

[fasmat]

Motivation

Part of #99. This change replaces the current proof generation with a new method using AES that should significantly speed up the process

Changes

• A new proving.Generate function has been added that implements the new algorithm for proof generation.
• A new verifying.VerifyNew function has been added that implements the new algorithm for proof verification.
• The old Prover has been marked as deprecated but not modified/removed to allow for benchmark comparisons
• The parameters for K1/K2, N, B have been chosen in a way that they work fast and reliably in unit tests. They are not optimised for security and performance in an actual network. For these parameters see: [Merged by Bors] - Integrate new PoST using AES go-spacemesh#4071

Optional:

• TODO: maybe first steps towards Provide a set of benchmarks / validations that will check the user platform and guesstimate how much storage could be commited #102 (postponed to Rust implementation)

[pigmej]

From logic perspective, from speedup perspective yes. But I did not checked the code ;-)

[pigmej]

Looked a bit more, @fasmat what's the reason for keeping the old implementation that we know it's broken?

[fasmat]

I already needed it multiple times for reference, like benchmarking and comparison with the current version. I wouldn't have been able to fix the system tests with the new version so quickly if the old version would not have been around any more.

The old algorithms are all marked as deprecated. If any code still uses them we will get a warning in the IDE and in our CI.

We can remove them at any point. Why does it have to be now, when we might still need them for reference?

[pigmej]

not only slow, @fasmat it's also broken and missing few items that would make it possible to generate proof correctly.

[fasmat]

Again, that's not the point. We are in a transitional period the "new code" is also already decided to be replaced with a new implementation in Rust.

It doesn't hurt to have them around for reference until we have the version we want to use for Genesis.