Empty issues should be empty arrays and not null

[timrobertson100]

As per this discussion on the SQL downloads I think we have started storing empty issue arrays in the parquet tables as NULL when previously they were empty arrays.

Should this be reverted?

(Note this probably applies to taxonomy issue arrays too)

[MattBlissett]

Yes, it makes sense to revert that. I'm sure we used to store empty arrays rather than nulls — it's documented as such for cloud parquet upload: https://github.com/gbif/occurrence/blob/master/aws-public-data.md