Documentation updates for predicate functions that rely on match

[nstepp]

Functions that rely on match and also take a function involving Context show expected, but poorly documented behavior.

The issue comes up when user functions that are supposed to operate on Contexts receive them as part of a recursive graph decomposition. As such, the graph that the contexts belong to gets smaller and smaller. User expectations are usually that if the starting graph has a particular context, that is the one that their function receives, as opposed to a context from a subsequent decomposition. This has come up before in #55 and #89.

I think this is at least a documentation fix, but it would also be useful to add some functions for mapping and folding over "full" contexts.

Would the maintainers be interested in pull requests for either?

[nstepp]

I just saw 84a89c4, but I still wonder if the ramifications of recursively calling match should be spelled out somewhere.

In any case, I'm also still curious about the possibility of something like cmap :: Graph g => (Context a b -> Context c d) -> g a b -> g c d or cfold :: Graph g => (Context a b -> c -> c) -> c -> g a b -> c that do not recursively call match. At the very least, it would be useful to have an efficient version of something that does Graph g => g a b -> [Context a b].