Add pyem-model-generator skill with templates, references, and workflow

[shawnrhoads]

Motivation

• Provide a reusable skill to scaffold new computational cognitive model modules and example notebooks compatible with pyem when requested models are not present in the base package.
• Support offline generation by codifying expected runtime contracts and parameter-recovery notebook patterns so generated code is consistent with the rest of the codebase.

Description

• Add a new skill under skills/pyem-model-generator including SKILL.md that defines behavior, mandatory clarification questions, generation workflow, and a smoke-check step to import the module and run sim/fit with output="npl".
• Add template.json as the primary scaffold template and add reference anchor templates rl.json, bayes.json, and glm.json that describe imports, signatures, transforms, and expected output keys.
• Add documentation assets README.md, references/pyem-runtime-contract.md, and references/parameter-recovery-notebook.md that specify utility contracts (e.g. softmax, norm2alpha, norm2beta, calc_fval) and minimal notebook/plot requirements for parameter recovery.
• Provide example generation targets and metadata such as parameter transform maps, fit/output contracts, and notebook sections (examples/{model_class}.ipynb path in the template).

Testing

• No automated tests were executed as part of this change; the skill includes a recommended smoke-check step but it was not run here.
• All new JSON and Markdown files were added as static resources and should be validated by downstream CI if present in the repository.

---

Codex Task