clutter: Align all screen transformations to 1/256th of a stage unit.
So as to eliminate floating point precision errors that creep in during matrix operations.
1/256th is chosen as a reasonable maximum resolution to cover any realistic fractional scaling factor. Floats can represent such a fraction losslessly because it is a small power of 2.
Closes #1403 (closed) (confirmed) and hopefully #1201 (closed). But see also 30809665 and !784 (closed) where the same kind of problem was encountered in the past.
Edited by Daniel van Vugt