ENH: improve granularity of performance tweak options for overlay operations

[theroggy]

The overlay operations now have performance-tweaking options via the batchsize and max_coords_subdivide parameters.

However, most overlay operations consist of several steps... that need different tweaks. It should be possible to pass specific tweaking parameters per step to be useful, as a batchsize for subdividing a layer is often not compatible with the actual overlay par example.