Revert cell detection multiprocessing to old version

[dstansby]

As a stopgap solution to #73 we (myself, @alessandrofelder and @adamltyson) decided to revert the multiprocessing logic in the cell detection to an older version that did not have the current memory issues, and put out a release.

I plan to:

1. Reproduce the memory build up locally on my laptop (done in Use random arrays for memory testing #82)
2. Revert to an older version of the multiprocessing logic
3. Make sure that the memory build up isn't an issue with the older logic

[dstansby]

Decided not to go forwards with this