Compare performance strategies for Dask interp_points

[rhugonnet]

See here:

geoutils/geoutils/raster/distributed_computing/dask.py

Line 294 in 209969f

# TODO 2: Check if computing block_i_id matricially + using an == comparison (possibly delayed) to get index

Moved TODO text:

Check if computing block_i_id matricially + using an == comparison (possibly delayed) to get index
 per block is not more computationally efficient?
  (as it uses array instead of nested lists, and nested lists grow in RAM very fast)

Also, using argsort() is usually pretty bad on performance, so could be good to test another option:

geoutils/geoutils/raster/distributed_computing/dask.py

Line 410 in 209969f

argsort = np.argsort(indices)