Explore: data-dict.yaml as the canonical metadata source (roxygen, JSON-LD, README derived from it)

[larnsce]

Motivation

Once #105–#107 land, washr would have two authoring surfaces: data-raw/dictionary.csv (canonical today) and data-dict.yaml (generated). Long term the pipeline could invert — authors edit data-dict.yaml (with the data-dict LSP/VS Code tooling) and washr derives everything from it:

• setup_roxygen() reads YAML and emits richer @format blocks (units, allowed values, ranges)
• generate_jsonld() maps YAML → schema.org variableMeasured (touches the Reconsider the dataspice metadata workflow: retire add_metadata() or fix its location and use? #87 dataspice reconsideration)
• dictionary.csv / attributes.csv become generated compatibility exports
• data-dict render / export-spec output embedded as a proper dictionary page on the pkgdown site
• glossary: becomes the machine-readable home for WASH domain vocabulary (JMP ladder levels etc.)

Background: dev/research-data-dict-integration.md, section "Option C".

Why backlog, not a release

This is a breaking workflow change for every existing openwashdata package, and the spec is still v0.1.0. Decision inputs before this graduates to a release milestone:

• Experience from the v1.2.0 exporter + CI validation on a pilot package
• Spec stability beyond 0.1.x, and whether an official R wrapper package appears
• Interaction with the Consolidate the FAIR layer into one auto-populated update_metadata() (lifecycle: experimental) #68 update_metadata() consolidation and Reconsider the dataspice metadata workflow: retire add_metadata() or fix its location and use? #87 dataspice decision