Move centrosome's "log" transformation into ImageMath

[bethac07]

Is your feature request related to a problem? Please describe. In 4.1.3, we added back the option to do centrosome's version of a "log transform" before thresholding, and we (and anecdotally others) have definitely found it useful. Not surprising, Lee wrote a lot of good stuff!

I don't recall if it ever was in ImageMath before and it got removed in the move to CellProfiler 3 and the "trying to remove non-standard methods" push we did as part of that (it's definitely NOT a standard log!) or if it was never there specifically at all, but it does do a useful thing that produces very different results than the things we currently offer (sidenote - are there any cases where log2 legacy DOES work well? Maybe only in cases where data is rescaled 1-to-greather-than-1?), so it feels like we could offer it explicitly as a transform as well. It would also be just one fewer link explicitly to the centrosome package, aka would help us a tiny bit down the road to CP5.

Additional context PDF of the jupyter notebook used to create these images