Enhancement of the blockwise function for CO3D data scaling

[adebardo]

Context:
To enable efficient processing of CO3D data, which consists of multiple gigabytes, we need to refine the blockwise functionality, particularly the apply section, to integrate our RANSAC filter. Additionally, several modifications must be implemented to optimize processing using our latest advancements in tiling and the implementation of a multiprocessing cluster.

Modifications to be made:

1. Replacement of the subdivision grid creation

   • Integration of the new tiling method.
   • Generation of coordinates and the required number of tiles for an efficient iteration in multiprocessing.

2. Implementation of reprojection in multiprocessing

   • Modification of the _preprocess_coreg_fit_raster_raster function to integrate multiprocessing from GeoUtils.
   • Use of the previously created tile grid.
   • Creation and management of a multiprocessing cluster.
   • Reprojection of data on the tiles.

3. Reading and aligning the reference DEM

   • Loading the DEM in tiling mode.
   • Processing in multiprocessing using the appropriate class.
   • Each worker will execute the .fit function on the tiles provided by the tiling function.

4. Replacing the call to stat with get_stats

   • Removal of the stat call in blockwise.
   • Replacement with get_stats, which is now available for each raster.