Map touch gesture differentiation should be done centrally and interpretation code overhauled

Currently we have duplicate code (or no code at all in some cases) for differentiation between different kinds of touch gestures on a map, sitting inside each different kind of mouse area or event. This is bad and leads to a number of issues, including pinches being interpreted as pinch + pan (giving incorrect results where the "zoom" feels like it's following the first touchpoint instead of the centroid of the gesture), flick+pans being interpreted incorrectly or not at all, and so on.

This needs to be overhauled, replacing the duplicated code with one shared gesture differentiator which then passes events on as appropriate (the parts of the differentiator can be co-located in each gesture's file, of course). Interpretation for events like pinch events needs to be overhauled so that it feels more appropriate in use.