Prepare v0.3 metadata-complete active index
Populate all 129 Pre-Standard rows with stable IDs, Xet file identities, byte sizes, and encoded-stream metadata; correct three canonical Xet IDs; split the Viewer into canonical/archive and pre_standard/archive; strengthen validation and release documentation. No video payloads are changed.
Verification summary (PR head f1f961606c50178f3f8aa78db73558668338f45b):\n\n- Dataset card parses and validates with the Hugging Face library.\n- Declared 29-feature order matches both CSV indexes.\n- Repository audit passes with 467/467 Xet IDs and byte sizes matching the pinned Hub resolve/inventory data.\n- Hugging Face Datasets 4.8.4 loads refs/pr/8 in streaming mode with video decoding disabled.\n- canonical/archive: 338 rows, 338 unique record IDs, 338 unique Xet IDs.\n- pre_standard/archive: 129 rows, 129 unique record IDs, 129 unique Xet IDs.\n- No video payload is changed by this PR.\n\nCurator decisions intentionally remain open for the 15 canonical 30x40 Paper vs. Illustration Board conflict, five unknown-date 00:00 values, and two incomplete Pre-Standard material descriptions.
Marking ready after all pre-merge validation passed.
Merging after local, Hub-identity, dataset-card, and refs/pr/8 streaming validation. The v0.3.0 tag will be created only after the production Viewer/Parquet build is verified.