Some color models support more tools, filters and other things like
colour selectors than other color models. Some support far less of
those things, in fact.

Among these things are:

* tools
* palettes
* filters
* paint ops

Thus, if a paint device is of a certain color model, certain GUI things
must be activated and deactived when that paint device becomes active.

A paint op may need to knwo something about the layer it is going to paint
on: it is not sufficient to generate a mask and have that composited by
the color strategy because the footprint may be determined by the deposit
and height field that is already present.

For some color models, pixels in a paint device must be
initialized using more or less complex algorithms. It is not enough to
initialize a single default pixel (which we cannot do yet), we must
additionally initialize the whole default tile; and since nothing in
Chalk outside the tilemanager code should know about the very existence
of tiles, we must find a generic solution of the canvas initialisation.

Additionally, some color models need permanently running filters to model
physical pocesses, like drying and flowing of paint or ink, or adsorbtion into
lower layers.

Finally, some color models (like the selection, or wetdreams) want a way to
efficiently add some kind of visualisation at the paintView level, instead
of the rendering level.