Track Poseidon2 benchmark baseline and degradation model for future hash selection

## Background

The roadmap defers hash-function replacement until ecosystem convergence, but still relies on:

- Poseidon2 as the current recursion benchmark baseline
- optimistic / middle / pessimistic degradation estimates for candidate replacements

## Scope

- Define the benchmark baseline that should be maintained until the hash decision
- Record the modeling assumptions for candidate replacements
- Make the output usable by recursion / prover planning
- Keep this issue decision-neutral with respect to the eventual replacement