Collaboration request: neural surrogate validation artefacts

[anulum]

Blocked modules

• src/scpn_control/core/neural_transport.py
• src/scpn_control/core/neural_turbulence.py
• src/scpn_control/core/neural_equilibrium.py

Needed evidence

• Immutable QLKNN, QuaLiKiz, gyrokinetic, transport, or equilibrium benchmark cases for cross-validation.
• Weight checksums, model provenance, feature scaling, prediction target definitions, fallback coefficients, collisionality convention, uncertainty metadata, and out-of-domain decisions.
• Matched reference equilibria or identical P-EFIT/EFIT cases for neural equilibrium.
• Error metrics for each validation case and a reproducible validation command.

Claim unblocked

Quantitative neural surrogate validation. Until these artefacts exist, tests only prove local loading, fallback, and regression behaviour.