Delete the dataspice set and trim the export surface to 9 (closes #100 when it ships)

[larnsce]

Rescoped 2026-09-02 per the #67 amendment (decision A in #111, triage #113). None of these exports has ever shipped on CRAN, so removal costs no deprecation cycle. The previous scope (fix update_access() and update_attributes()) is moot because the functions go.

Tasks:

• Delete add_metadata(), add_creator(), update_access(), update_attributes() and update_biblio(), with their R files, Rd files and tests. update_gsheet_metadata() goes in Remove update_gsheet_metadata() from the package #69.
• generate_jsonld() becomes internal (the builder behind update_metadata(), update_metadata(): derive the schema.org JSON-LD from the canonical sources, embed it in the pkgdown head, report blanks (absorbs #70) #68). Unexport fill_dictionary() and generate_roxygen_docs().
• Remove the data/metadata handling (the .Rbuildignore entry and the folder scaffolding). Existing packages keep their folders and washr ignores them.
• Update the _pkgdown.yml reference index ("Adding metadata" block) so the site builds.
• NEWS entries for the removed exports and for the changed update_metadata().
• The PR that ships this carries "Closes update_access() breaks for multi-dataset packages: repo URL from dataset name, recycled encodingFormat, assumed csv+xlsx pairs #100"; the update_access() bug disappears with the function.

Acceptance: NAMESPACE contains exactly the stable core seven, use_brand() and update_metadata() (9 exports, down from 18); R CMD check clean; the pkgdown reference index builds.

Milestone: v1.1.0.

[larnsce]

Triage note from #113 (2026-09-02). Two changes that do not depend on any decision, and one that does.

• The acceptance line now lists use_brand() next to the stable core seven. It shipped on dev through use_brand(): install and refresh the openwashdata brand #110 and use_brand(), the #67 decision document, fixes for #101 and #102 #114, so the export count this issue trims from is 18, not 17.
• update_access() breaks for multi-dataset packages: repo URL from dataset name, recycled encodingFormat, assumed csv+xlsx pairs #100 adds one point to the update_access() task: enumerate inst/extdata instead of assuming a csv plus xlsx pair per dataset, so a csv.gz export gets a row and no row is invented for a file that does not exist. The repo segment and the recycling bug in update_access() breaks for multi-dataset packages: repo URL from dataset name, recycled encodingFormat, assumed csv+xlsx pairs #100 are already in the task list.
• The scope of this issue depends on the Design note: washr as a deterministic engine with a Claude skill as driver (review of exports and open issues before the next step) #111 decision (decision support is posted there, answer due 2026-09-16). If Metadata design decision: canonical sources and field mappings (resolves #47) #67 stays as agreed, this issue stands as written plus the update_access() breaks for multi-dataset packages: repo URL from dataset name, recycled encodingFormat, assumed csv+xlsx pairs #100 point. If Metadata design decision: canonical sources and field mappings (resolves #47) #67 is amended as Design note: washr as a deterministic engine with a Claude skill as driver (review of exports and open issues before the next step) #111 proposes, the update_access() and update_attributes() tasks are moot because the whole dataspice set is deleted, and this issue becomes: delete add_metadata(), add_creator(), update_access(), update_attributes(), update_biblio() and update_gsheet_metadata(); make generate_jsonld() internal; unexport fill_dictionary() and generate_roxygen_docs(); update the reference index; NEWS records the removals (none ever shipped on CRAN). Exports go from 18 to 9, and update_access() breaks for multi-dataset packages: repo URL from dataset name, recycled encodingFormat, assumed csv+xlsx pairs #100 closes with this issue.