Updates on metadata

new splits w radec footprints

changed csv to jsonl

new splits with removed data. 500 cubes are left

Created new train/test splits for SDSS which ensure that there is no overlap between train and test sets

add dataloader and splits