Generic traversals

[gampleman]

This somewhat superseeds #17, as this adds more powerful, generic traversals, which makes the need for too many variations on map less necessary, since anyone can quite easily make their own.

In fact these are so powerful, that I refactored the whole library to sit on top of these. I think this has the following benefits:

• smaller code size, as the core bits that do tail recursive traversal exist fewer times
• potentially in applications there is a higher chance that since the same code is excersized more, it will get JITed to a higher perf category
• smaller surface to introduce bugs into

On the negative side there is potentially some small overhead as there is a bit extra information being ferried around that isn't strictly necessary for all the functions involved. My hope is that a) this is fairly negligible and b) that the aforementioned higher probability of JIT will pay for it in practice.

However, in a 1-to-1 microbenchmark, it's likely some methods would slighly regress. However, they may perform better in real world applications.