Matsya DP workflow and Dolo draft for GKN HANK model

[Calbafa]

Session identifier: topics2026-mon-policy

This pull request implements the Matsya-based dynamic programming workflow for the Gornemann, Kuester, and Nakajima (2012) heterogeneous-agent monetary policy model.

The workflow was executed using a single stable Matsya session across all queries, ensuring that context was preserved and can be reviewed on the server side.

This PR includes the following artifacts:

• Bellman excerpt markdown (docs/gkn-bellman-excerpt.md): a concise and faithful summary of the household dynamic programming problem, including states, controls, constraints, and Bellman equations.
• Improved markdown (docs/gkn-bellman-improved.md): a refined version of the model description using a stage-based structure inspired by SolvingMicroDSOPs.
• Dolo Plus draft (docs/gkn-dolo.yaml): a simplified YAML representation of the model constructed incrementally using Matsya.
• Verification paragraph (docs/gkn-verification.md): a short discussion comparing Matsya outputs to the original paper, including accepted elements and necessary adjustments.
• Session identifier file (docs/matsya-session.txt): records the Matsya session used for all queries.

Due to the complexity of the full HANK model, the YAML representation abstracts from certain features such as the full distribution of agents and general equilibrium feedback. These simplifications are documented in the verification file.

Matsya was used iteratively to:

1. Decompose the Bellman problem into stages,
2. Clarify timing and expectation structure,
3. Build a structured YAML draft.

All Matsya calls were executed with the same session identifier to ensure continuity and reproducibility.