Define the term as 'schema dimension' (shortened to 'dimension')

[dimitri-yatsenko]

Promotes schema dimension to the defined term, then shortens to dimension in-context.

Why

dimension is overloaded — data-warehouse dimension tables, OLAP cubes, and array/tensor dimensions (the type-system and npy pages already use "dimension" for array axes). Qualifying it as schema dimension on definition disambiguates cleanly. This isn't a new coinage: the Entity Integrity page already titles the section ## Schema dimensions and says "lineage through schema dimensions" — this just makes the defined term consistent with that.

Changes (definitional homes only)

• Entity Integrity → §Schema dimensions: A **dimension** is… → A **schema dimension** — often shortened to *dimension* — is…
• Read Diagrams → §Dimensions and Underlined Names: the same promotion on the definition sentence.
• Everything else keeps the short form dimension (define once, shorten in-context) — so all headings and anchors are unchanged.

Notes

• Overlaps only lightly with the WIP diagram PR docs: diagram operations — fold cascade/trace/restrict derivation into the specs (WIP) #234 (scoped to diagram operations — cascade/trace); different sections, low conflict.
• llms-full.txt regenerates from source (not hand-edited here).
• The Read Diagrams cell source was normalized to a single JSON string by the notebook editor (valid nbformat, identical rendering).

[MilagrosMarin]

The dimension → schema dimension promotion is well-motivated (the collision with warehouse dimension tables and array axes is real, and the Entity Integrity page already titled the section ## Schema dimensions), and the "define once, shorten in-context" approach keeps all headings and anchors stable. CI green.

Three mechanical notes on the notebook, none about the wording itself:

1. Trailing newline dropped — the diff ends with \ No newline at end of file. Every other notebook in the repo (model-relationships, 02-schema-design, master-part, and read-diagrams on main) ends with one, so this is the only file without.

2. The cell-source normalization is worth undoing. You flag it as benign nbformat, and it is — but on main all 54 cells in read-diagrams.ipynb use list-of-strings source; on this branch that cell is now the only plain-string one. The practical cost is that any future edit to that Markdown cell diffs as a single unsplittable line rather than per-line — which matters for a cell likely to be touched again.

3. Parenthetical style differs between the two definitional homes — entity-integrity.md uses em-dashes (— often shortened to *dimension* —), the notebook uses parens ((often shortened to *dimension*)). Same meaning, just reads as drift given both land in this PR.

Approving — none of these are correctness issues.