Fix: Use __name__ instead of custom NAME attribute for class registration

[aditya0by0]

This PR updates the model registration logic to use Python's built-in cls.__name__ instead of a custom cls.NAME attribute when checking and storing class references in _MODEL_REGISTRY.

🔍 Reason for the change:

• Using cls.__name__ is more robust and Pythonic—it ensures every class is uniquely identified by its actual class name without relying on the presence or consistency of a manually defined NAME attribute.
• Prevents potential issues where cls.NAME may be missing, misdefined, or inconsistent across subclasses.

This should make subclass registration safer and eliminate silent bugs due to missing or duplicate NAME attributes.

[sfluegel05]

Can you tell me why we need this name in the first place? It is saved to a _MODEL_REGISTRY, and then? Is this registry ever used?

[aditya0by0]

Can you tell me why we need this name in the first place? It is saved to a _MODEL_REGISTRY, and then? Is this registry ever used?

The model registry is used to prevent duplicate subclasses extending ChebaiBaseNet—whether they’re defined in the same repo or in an external package like chebai-graph that builds on top of chebai.
Previously, this check relied on each subclass having an explicit NAME attribute, but this PR replaces that with cls.__name__ to make it safer and less error-prone.
Whenever a new subclass is created by extending ChebaiBaseNet, it gets registered in _MODEL_REGISTRY, so yes, the registry is actively used to ensure unique model definitions and avoid silent conflicts.

[sfluegel05]

Allright, then let's keep it that way.