MLTheory proof chapter

This is the curated entrypoint for theorem-heavy MLTheory developments. It groups the proof files by mathematical theme:

• approximation and finite-precision universal approximation;
• Hopfield energy descent and convergence;
• ReLU algebra and compact-set approximation;
• state-space / Mamba scan and causality laws; and
• verification-facing robustness theorems.

We keep this as a proof chapter rather than mixing it into model definitions. The model/spec layer defines semantics; this chapter proves reusable mathematical properties about those semantics.