aggregate_to_distribution yields mulitpolygons with new refactor output

[mikejohnson51]

I am sure this does not have to do with the refactor output but exposes a gap in the aggregation logic:

Red = uniform output Tan = input catchments 10015624, 10015623, 10015622, 10015986, 10015985, 10014167 (main offender: 10014168)

[mikejohnson51]

Occurs in collapse_headwaters2 call (as expected). I gets merged into the catchment where water is "flown" rather than where the junction occurs.

[mikejohnson51]

closed with #16