macrostrat.units max_thick and min_thick should allow NULL

[cambro]

It is essential that at least min_thick allow NULL values to distinguish meaningfully from zero values. This was a major design flaw on my part from the start.

Currently, min_thick=0 can mean a unit actually has a minimum of 0 m or min_thick was not specified and defaulted to 0 m, which is definitely not the same thing. So, we have a big ambiguity problem of my creation. We can fix future data now.

[davenquinn]

The NOT NULL constraints for these columns were autogenerated by the MariaDB -> PostgreSQL translation process. I agree that they should be removed and these fields should be null when not defined.