#125 Add partial identification and bounds lecture

Continues the assumption-focused causal-inference layer.

What changed

• adds 04-causal-inference/partial-identification-bounds/;
• adds the canonical presentation/main.tex entry point;
• adds Partial Identification and Bounds;
• covers point vs partial identification, identification regions, worst-case/Manski bounds, sharpness, monotonicity restrictions, IV bounds, Lee bounds, missing-outcome bounds, sensitivity regions, and set-valued inference;
• registers the deck in the presentation contract validator;
• registers the deck in GitHub Actions and compile_all.sh;
• updates the causal-inference domain index.

Review note

This PR intentionally does not merge automatically.

---

Summary by cubic

Adds a new lecture deck on partial identification and bounds to the causal-inference curriculum, introducing set-identified causal conclusions as an alternative to point identification.

• Covers Manski worst-case bounds, sharp versus non-sharp bounds, monotonicity restrictions, IV/Balke–Pearl bounds, Lee bounds for missing outcomes, sensitivity regions, and set-valued inference.
• Registers the deck in the GitHub Actions LaTeX workflow, compile_all.sh, and the presentation contract validator, and adds it to the causal-inference domain index.
• This PR intentionally does not merge automatically.