This massively reduces the number of temporary ol.TileCoord objects.
Previously an ol.TileCoord object was generated for every potentially
visible tile at the current zoom level and lower, every frame. This
commit eliminates all of those. Now new ol.TileCoord objects are only
allocated when a new tile is created.
With this commit we register a change listener at a time when we are guaranteed that the listener will be called. In this was we don't need to remove listeners when the tile is dropped from the tile queue. And to avoid busy-polling between IDLE and LOADING we rely on the map to call requestRenderFrame when new tiles change to LOADING from IDLE.
This change adds a lot of flexibility to working with tile
layers: Sources where the server projection or tile grid do not
matter can now be constructed without specifying a projection or
tile grid.
The tileUrlFunction/imageUrlFunction now also creates updated
URLs when the params of the layer change, so things like
mergeNewParams in ol2 will be possible.
A nice side effect of this whole change is that there is no more
duplicated code between tiled and single image WMS layers.
While I was at it, I also fixed a WMS 1.1.1 axis order issue
and incorrect STYLES params (STYLES=& instead of STYLES&).
Instead of going into an unnecessary animation loop, we can simply wait to be notified when tiles load.
The WebGL renderer still sets frameState.animate true, but I think this too should be unnecessary (full page example works without it, side-by-side example shows it cannot yet be removed).
Instead of keeping track of wanted tile ranges, we can instead track wanted tiles individually. This provides enough for the map to know how to prioritize tiles and should be more efficient (no extra calls to extend tile ranges or check for tile containment within a range).
