[DO NOT MERGE] decaf version bumps decaf

[TalDerei]

Experimenting with #3676

[TalDerei]

continuing debugging proof failures

[TalDerei]

Currently, proof verification is still failing. There are a few noteworthy observations to mention:

• Old backend (v0.5) generates and verifies proofs with actual prover / verifier keys and dummy keys,
• New backend (v0.8) generates and verifies proofs with dummy keys, but not with actual prover / verifier keys,
• The representation of the dummy keys, for instance for a spend circuit, for some reason differs between v0.5 and v0.8.

Interestingly, in v0.8, the blinding factors, pk, and public and private inputs in spend_proof_parameters_vs_current_spend_circuit match up against the old backend, but proving generates a different proof result than expected and fails proof verification for spend proofs. Simply swapping the actual proving keys for the dummy keys succeeds.

As an aside, I generated a proof with v0.5, and verified the proof string directly in v0.8 which passes the verification checks with the actual verification key. This suggests there may be an issue instead with how we're serializing specifically our proving keys.

@redshiftzero I created a minimal crate called old-backend with calls to generate_constraint_matrices and generate_test_parameters which pass, which suggests our new backend is properly creating R1CS matrices.

[TalDerei]

Closing in favor of updated #3806