Changes in version 0.5.0                        

  - Added the 2004–2021 Alvarás deposit as alvaras_sp_2004_2021,
    alongside the existing 2010–2024 alvaras_sp.
  - Updated get_dataset(docs = TRUE) to return dictionary workbooks for
    datasets that provide them.
  - Added query_dataset(), which opens a parquet dataset as a lazy
    DuckDB table that reads from Dataverse without downloading the whole
    file.
  - Changed the registry to load from GitHub once per session, so
    datasets added after a release appear without reinstalling. The
    package falls back to its shipped copy when GitHub cannot be
    reached, and options(inspercidados.registry = "bundled") forces the
    shipped copy.

                        Changes in version 0.4.0                        

  - Added list_resources(), which lists the logical datasets, years,
    formats, and defaults within a registered deposit.
  - Added validation for malformed get_dataverse() file selectors.
  - Changed get_dataset() to select a logical resource before choosing a
    file format. This is a breaking change.
  - Moved raw filename and file_pattern selection from get_dataset() to
    get_dataverse(). Unregistered DOIs now go through get_dataverse().
  - Renamed browse_project() to open_project() to avoid a conflict with
    usethis::browse_project(). browse_project() still works but warns.
  - Deprecated get_script() in favor of open_project(), since pipelines
    are now published as repositories.