ImageMath types

[amnipar]

Some ImageMath functions have type issues that are revealed by haddock documents. Like, findMinMaxMask uses BareImage. It is also dubious which images can (or should) be called with D8 images. In my documentation commit, I have already changed some functions (like sqrt) to accept D32 image only. Then addS is now for D32 only while it might be possible to use it for D8 - but there could be overflow problems. Suggest deciding this matter for every function.

As a side note, should we have a BinaryImage type (perhaps newtype of Image GrayScale D8) that would be the only image type accepted for masks and such, and could be produced by comparison operators?

[aleator]

Just an update: Could you be more specific on what needs to be changed?