Rewrite LP lecture: drop OR-Tools, focus on SciPy, add duality (#514)

Full rewrite of the Linear Programming lecture addressing the editorial
suggestions in #514.

- Drop OR-Tools (GLOP) entirely; solve everything with scipy.optimize.linprog
  (HiGHS), removing the ortools dependency and the dual-solver presentation.
- Restructure the flow: pose both examples first (Example 1 solved
  graphically, motivating the need for a systematic method on the
  higher-dimensional Example 2), then introduce standard form, then solve
  both with SciPy.
- Add a Duality section: derive the dual of the production problem,
  state weak/strong duality, interpret the dual variables as shadow
  prices, and verify the material shadow price by perturbation.
- Note why we use real-valued (vs integer) decision variables.
- Rewrite both exercise solutions using linprog.

All code cells verified to execute via jupytext.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: jstac