Cells should be assignable to multiple regions if they fall into two segmentation regions

[ambrosejcarr]

trafficstars

Currently Target assignment overwrites the cell_id for a feature if it is found in more than one cell region. However, we should not do that:

@berl made a critical observation about cells with multi-region membership here: https://github.com/spacetx/starfish/pull/368#pullrequestreview-144104105

[berl]

@ambrosejcarr sorry if I'm being thick here, but this issue title is difficult to comprehend. The concern you referenced was that spots should be assigned to multiple regions if they're segmented as such. Is your title a different way of saying that?
One design to address this problem would be to leave the IntensityTable at this point and move to a CellTable wherein each cell could have whatever attributes it needs, including spots that were shared with another cell. No such structure exists now, right?

[ambrosejcarr]

Sorry I just moved a comment of yours out of a larger issue that was mostly closed. And I didn't make the title that clear.

[neuromusic]

For now, we won't be supporting this