I-diagram matrix: connected hollow/filled dots and Jurgens atom labels

[Autoplectic]

Summary

• In the UpSet dot matrix of plot_information_diagram, every atom column now has a single connecting line spanning all variable rows, instead of a line spanning only the included variables.
• Included variables are drawn as filled dots in the atom's colour; excluded variables are hollow dots (white face, atom-coloured edge) so the line doesn't show through them.
• Each dot chain is labelled underneath with the atom's taxonomy name from Jurgens & Crutchfield (2026) Table II (t.rμ, p.r⁺μ, …, qμ, σμ) via IDiagramAtom.jurgens_label. The seven cancelling-partner atoms not in Table II keep their existing † labels, and always-zero atoms (only shown with atoms="all") fall back to the set-builder label.
• Default figure height grows from 5.6 to 6.4 in, with a fixed ~0.85 in bottom strip for the rotated labels.

The 14 Table II labels in JURGENS_LABELS were checked against both the Taxonomy of Prediction and Information Machines manuscripts; they agree.

Test plan

• uv run pytest tests/test_information_diagram.py (45 passed)
• Rendered nemo with default and atoms="all" and checked the matrix and labels visually