Set up MkDocs documentation system with mkdocstrings

[sidhantpande]

Summary

Following up on #61, this issue proposes setting up MkDocs with
mkdocstrings as the documentation system for neural-lam.

What this adds

• mkdocs.yml config at the repository root
• docs/ directory structure with:
  • index.md (project overview)
  • getting_started/installation.md
  • getting_started/quickstart.md
  • api/metrics.md (auto-generated from docstrings)
  • api/models.md (auto-generated from docstrings)
  • api/datastore.md (auto-generated from docstrings)
• MkDocs + mkdocstrings added as optional docs dependency group
  in pyproject.toml

Why MkDocs over Sphinx

• Simpler configuration (YAML vs Python)
• MkDocs Material theme is widely used in Python projects
• mkdocstrings auto-generates API reference from NumPy-style
  docstrings already present in the codebase
• Easy GitHub Pages deployment via CI

Related

• Structured documentation for Neural-LAM #61 (Structured documentation for Neural-LAM)
• GSoC Project Idea Area definition as part of static dataset files #2: Automatic generation of documentation

[sadamov]

please discuss this directly in #230 / #61 I am not evaluating MkDocs vs. Sphinx at the moment. But we don't need multiple PRs or issues for #61. If you don't agree with the solution proposed in #230 you should present your case for MkDocs directly there.