Speed up fogpy

[gerritholl]

Fogpy is rather slow. Needs some profiling, and probably some of the lower-end functions could be rewritten in numba to speed things up considerably.

[gerritholl]

https://medium.com/capital-one-tech/dask-numba-for-efficient-in-memory-model-scoring-dfc9b68ba6ce

[gerritholl]

I expect that instead of looping through all pixels such as done here, it should be possible to loop through all objects (see also scipy.ndimage.find_objects).

https://github.com/gerritholl/fogpy/blob/2c9aa9436e4924f55879d52a9de14817d370df5a/fogpy/algorithms.py#L644-L671

[gerritholl]

Need to check if a small region is faster than a large region. It should be. If it isn't, something with the order of resampling is not correct.