L_2: Introduce metadata part in .vnnlib file to persist image, image_class and epsilon for SDP-CROWN

[henba1]

Introduce VNNLib sidecar for SDP-CROWN that carries image and image_class.

For SDPCrown, carrying the image alongside the VNNLIB property is the cleaner and less error‑prone design than trying to reconstruct a BoundedTensor inside SDP-CROWN from the VNNLIB text.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Aaron99B]

The image is already included in the verification context in the data point. Why do we need it another time?

[henba1]

The image is already included in the verification context in the data point. Why do we need it another time?

VerificationContext is VERONA-internal state by design of the current auto_verify_module.py, which from verification_context only passes along verification_context.network.path to the verifier via verify_property(...).

Recall that auto-verify verifier API is
def verify_property(
self,
network: Path,
property: Path,
*,
config: Configuration | Path | None = None,
timeout: int = DEFAULT_VERIFICATION_TIMEOUT_SEC,
) -> CompleteVerificationResult:

But SDP-CROWN needs image, image_class/label and epsilon. Thats why we attach it to VNNLib_property

[Aaron99B]

Can't we include SDP-Crown using a different interface? I would refrain from putting the same image twice in the verification context