Expose independently selectable Standard recording workflows
Researchers currently enter both Standard recording workflows through the combined canonical configuration. This change adds standard_earlier (364 files) and standard_individual_recordings (353 files), generated solely from existing acquisition_protocol_version assignments. canonical remains the default; pre_standard and the archive split are preserved.
Each new native configuration uses the root metadata, its exact selected media paths, and a protocol predicate. Generated CSV exports preserve all 35 source columns and every cell value. The individual-recording selection includes all 175 known and 178 unknown starts; timing availability and unresolved drawing identity do not control membership.
The supplied implementation is integrated into the existing manual ingest procedure after the identity/timing projections and canonical count update. Drift checks run before publication. Windows fixture newline and path-redirection compatibility are fixed; native checks additionally verify the exact resolved input paths, and full-repository native loading is included in regression tests.
Validation at parent c3e959c8c7b4f60f19b34c5501b19459269d53a5:
- 150 repository tests passed; zero skipped.
- Named native loading with datasets==4.8.5 passed: 364 / 353 rows, all fields and media references preserved, zero video payload reads.
- Disjoint/exhaustive membership and byte-idempotent rebuilding passed.
- Both source indexes, media, checksum manifest, work registry data, relationship/timing evidence and frozen releases are unchanged.
- Nine changed files, including current maintainer documentation; no schema changes, media changes, tags, or GitHub browser changes.
A revision-pinned metadata-only selection example saves exact record IDs before downloading videos. This PR is not a merge or proof of hosted Viewer processing. Hosted configuration counts, sample IDs and revision headers remain a separate post-merge check.