Build MVP datasets: CSV sources in data-raw → exported data objects + Parquet

[dlebauer]

Goal

• Add the MVP datasets to the package so users can load them as exported data objects (per r-pkgs data guidance),
  with CSV sources versioned in git and Parquet artifacts for distribution/performance.

Scope (MVP)

• Primary exported dataset: traitsview (packaged name for traits_and_yields_view).
• Supporting metadata tables required to interpret records (variables, sites, species, citations, treatments, cultivars,
  entities, methods, PFT mappings, priors). Exact table list is already known from prior scoping; do not re-argue it.

Tasks

• Place editable sources under data-raw/csv/:
  • data-raw/csv/traitsview.csv (traits_and_yields_view exported as CSV or recreated with joins)
  • data-raw/csv/<support_table>.csv (support tables)
• Create data-raw/make-data.R that:
  • reads CSVs (stable column typing)
  • writes exported data objects via usethis::use_data(..., overwrite = TRUE)
    • exported objects: traitsview, plus each support table as its own object (e.g., variables, sites, etc.)
  • writes Parquet copies under inst/extdata/parquet/ (e.g., inst/extdata/parquet/traitsview.parquet)
• Ensure exported objects load with library(betydata); head(traitsview).

Acceptance criteria

• After installation, users can do:
  • library(betydata); head(traitsview)
  • library(betydata); head(variables) (and other support tables)
• data-raw/csv/ contains the CSV sources committed to git.
• inst/extdata/parquet/ contains Parquet artifacts for all shipped datasets.
• A clean rebuild path exists: running source('data-raw/make-data.R') regenerates data objects and Parquet.

Non-goals

• No accessor-function API (avoid bety_tbl() style).