Make streaming schema deterministic
Repair the Pre-Standard index for clean Hugging Face Datasets loading: remove the entirely empty original_filename column that can infer as float/null against Standard string values; add collection_tier=pre_standard and is_canonical=false to all 129 rows; and document a bandwidth-safe streaming quick start plus the optional torchcodec decoding requirement. All existing descriptive values and media paths are preserved.
FINAL REVIEW SCOPE: align both metadata.csv files to the same 29-column contract; preserve all 129 Pre-Standard media paths and seven existing descriptive fields; populate only collection_tier=pre_standard and is_canonical=false; keep undocumented fields empty/null; declare exact dataset_info types; and add a bandwidth-safe streaming Quick Start with the torchcodec decoder requirement. Verified with Hugging Face Datasets 5.0.1 in three clean passes: 467 unique rows, 338 Standard, 129 Pre-Standard, uniform schema, and no video transfer. Only README.md and Pre_Standard_Time_Lapses/metadata.csv change.
Merged after three clean Hugging Face Datasets 5.0.1 streaming passes: 467 unique rows, 338 Standard, 129 Pre-Standard, exact 29-field schema, preserved Pre-Standard descriptive values, and no video transfer.