Use `pool` to download items with `filter`

[manycoding]

The current logic of dividing on batches by start_index and count doesn't account for filter. When using filter, returned items _key don't correspond with actual index so the data repeats.