Add IFP II lecture on Optimistic Policy Iteration

- New ifp_opi.md lecture implementing OPI for income fluctuation problem
  - Uses Model NamedTuple structure consistent with ifp_discrete.md
  - Implements policy operator T_σ and iterate_policy_operator
  - Provides comprehensive timing comparisons between VFI and OPI
  - Shows 2-2.5x speedup from OPI over VFI
  - Includes exercise exploring parameter sensitivity

- Update _toc.yml to include ifp_opi between ifp_discrete and ifp_egm

- Update ifp_egm.md title from "IFP II" to "IFP III"
  - Reflects new lecture ordering with OPI as second lecture

Technical highlights:
- Uses @jax.jit decorators and jax.lax.while_loop for performance
- Properly handles JAX traced values with jnp.where instead of Python if
- Tests multiple values of m (policy iteration steps) to find optimal
- Visualization comparing OPI performance across different m values

Tested: Converts to Python and runs successfully with realistic speedups

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: jstac

lectures/_toc.yml

@@ -85,6 +85,7 @@ parts:
   numbered: true
   chapters:
   - file: ifp_discrete
+  - file: ifp_opi
   - file: ifp_egm
   - file: ifp_advanced
 - caption: LQ Control

lectures/ifp_egm.md

@@ -19,7 +19,7 @@ kernelspec:
 </div>
 ```
 
-# {index}`IFP II: The Endogenous Grid Method <single: IFP II: The Endogenous Grid Method>`
+# {index}`IFP III: The Endogenous Grid Method <single: IFP III: The Endogenous Grid Method>`
 
 ```{contents} Contents
 :depth: 2