The PDF build on ReadTheDocs produced empty content because
sphinx_selective_exclude.eager_only discarded the `.. only:: latex`
toctree at parse time, and RTD's shared doctree cache preserved
that decision for the subsequent LaTeX build.

- Remove eager_only extension and sphinx-selective-exclude dependency
- Merge HTML/LaTeX toctrees into a single unconditional toctree
- Rename inspecting-models to 14-inspecting-models for consistent naming
- Remove non-existent 12-state-estimation from toctree
- Add iframe_figures/ to .gitignore (myst-nb Plotly build artifact)
- Update v1.10.1 release date to 2026-03-09