feat: add GERD by sector of performance NUTS2 (RD_E_GERDREG) (#79)

[Gabrymi93]

Type

• Root fix (removes the cause of a bug)
• Contract alignment (aligns behavior across components)
• New dataset (adds a Eurostat dataflow to the registry)
• Infrastructure (CI, GCS, MCP, docs)
• Chore (cleanup, test, refactor)

Problem

First innovation/R&D dataset in the catalog — RD_E_GERDREG (GERD by sector of performance and NUTS2 region) was listed in the planned section but not implemented.

Solution

• Added sectperf codelist (28 codes: BES, GOV, HES, PNP, TOTAL) to scripts/update_codelists.py
• Created datasets/eurostat-rd-expenditure-nuts2/ with 4 dimensions (freq, sectperf, unit, geo)
• clean.sql enriches sectperf with label_en via codelist
• Pipeline: 909,090 clean rows → Italy NUTS2 mart

Key findings (2023, % of GDP)

Region	R&D spending
Piemonte	2.09%
Emilia-Romagna	2.09%
Lazio	1.80%
Calabria	0.64%
Valle d'Aosta	0.53%

Gap between top and bottom: 4×.

Test plan

• pytest tests/ -v passes (20/20)
• toolkit run full --config datasets/eurostat-rd-expenditure-nuts2/dataset.yml --years 2026 passes

Downstream impact

• Parquet schema changed (columns renamed or removed)
• GCS paths changed
• MCP tools affected
• None — new dataset only

Closes

Closes #79

Checklist

• ruff check connectors/ tests/ passes
• docs/dataset-registry.md updated
• docs/contributing.md template still accurate