CI validation of dictionary against data with the data-dict CLI

[larnsce]

Motivation

Today nothing checks that data-raw/dictionary.csv still matches the data — a renamed column, a type change, or a new variable only gets caught by eye. With data-dict.yaml (#105) and parquet exports (#106) in place, the data-dict CLI can enforce dictionary–data consistency mechanically:

• validate-spec — dictionary conforms to the YAML schema
• validate-meta — column names/types match the parquet files
• validate-data — values satisfy constraints (ranges, enum values, required, assertions)

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

Proposal

• Add a validation step to the workflow templates washr scaffolds (builds on the R-CMD-check scaffolding shape of Scaffold the R-CMD-check workflow so washr packages pass the review floor by construction #86): install the CLI (one-line curl installer or uvx data-dict from PyPI) and run the three validate commands
• Optional local companion check_datadict() that looks for the binary on PATH and skips politely when absent — washr is on CRAN and cannot hard-depend on a Rust binary; this also fits the check_publication_readiness() gate design in check_publication_readiness(): workflow status gates (port fairenough's validate_* design) #82

Acceptance criteria

• Scaffolded workflow fails CI when dictionary and data drift apart (renamed column, out-of-range value)
• Runs green on a pilot openwashdata package
• No hard dependency on the CLI for washr itself (CRAN-safe)

Caveat

The spec is v0.1.0 and may change; pin the CLI version in the workflow template.