Generate partitions/subsets of knowledge split by ioos_category

[abkfenris]

Add utils/generate_partitions.py (uv script) that reads core/standards/*.yaml
and emits both compact JSON and human-readable YAML for:
data/all-standards.{json,yaml} — full CF vocabulary
data/all-knowledge.{json,yaml} — CF standards + all knowledge entries
data/partitions/{category}.{json,yaml} — 14 per-IOOS-category self-contained subsets

Each partition: { cf_standards: {standard_names, aliases}, knowledge: [...] }
Consumers call loadStandards(data.cf_standards) + loadKnowledgeObjects(data.knowledge).

YAML scalars that look like numbers (e.g. 'unit: 1', 'unit: 0.001') are coerced
to strings in both scripts — matching serde_yaml's automatic coercion but
preserving correctness through the JSON path.

Extend gen-data.mjs to mirror the same partition logic into public/data/ for
Vite dev/test serving (identical format, JS-only dependency chain).

Add 3 Vitest tests verifying partition loading:

• meteorology subset is self-contained (out-of-category standards absent)
• all-knowledge covers every category
• all-standards loads the full vocabulary without knowledge enrichment

Add generate_partitions nox session (off by default, opt-in for CI release path).
Full nox suite passes: 19+11 Rust, 15 Vitest, 5 Playwright, 1 pack smoke.

---

This is part 4 of 6 in a stack made with GitButler:

•  6  Release artifacts #182
•  5  Use the knowledge partitons to make a NPM data package #181
•  4  Generate partitions/subsets of knowledge split by ioos_category #180 👈
•  3  WASM/JS external ingestion API #179
•  2  Feature gate for embedding data #178
•  1  JS build optimization #177

[github-actions]

Javascript size report

	Size
Base (phase-2b-wasm-js-ingestion-api)	462.3 KB
PR	462.3 KB
Delta	+0.0 KB (+0.0%)

✅ No size change.