The dataset_meta keys that live in the descriptive sidecar, not the notebook
Source: R/catalog_datasets.R
dataset_meta_descriptive_keys.RdPlan § D-9 (Decision 14) splits an ingest's calcofi.dataset_meta in two:
the structural keys stay in the notebook YAML (dataset_name,
dataset_name_short, category, color, tables, in_release — what the
pipeline needs to run and group), and the descriptive keys — the ones a
provider edits in a Google Sheet — move to
metadata/{provider}/{dataset}/dataset_meta.yml. read_calcofi_meta()
merges the two; scripts/build_workflows_index.R errors when a descriptive
key is still found in a notebook.