Add overview support for imports

[migurski]

• Reduced dual GDAL pipes to a single one in each script
• Added calculation of overviews for all zoom levels back to z=0 using a recursive descent to keep only a small window of the original raster in memory

Part of #8

[migurski]

Sample overviews of test file tests/Annual_NLCD_LndCov_2023_CU_C1V0.tif using nearest-neighbor resampling:

[cayetanobv]

It's working; several comments only:

• minresolution and overviews created. We're setting minresolution=0 because we're computing all overviews, which is not useful in all use cases. It's better to follow the GDAL strategy, computing the minresolution overview depending on the geographical area covered. More comments below in the code.
• Stats: we should support approximate stats. In large rasters to compute exact stats is a huge overhead, and probably you can work fine with approximate stats when the purpose is visualization. The approach in this PR is computing stats block by block, which is memory safe, but it could be slow for big rasters.
• Weird issue: The colointerp capitalized is not interpreted correctly in CARTO (GDAL returns as is included here). The colortable is not used if the optional stat "top_values" doesn't exist. We can talk about that because this is something specific to CARTO.

[migurski]

• Stats: we should support approximate stats. In large rasters to compute exact stats is a huge overhead, and probably you can work fine with approximate stats when the purpose is visualization. The approach in this PR is computing stats block by block, which is memory safe, but it could be slow for big rasters.

Do you recommend that we need to do it in this PR, or is this a suggestion for future work?

[cayetanobv]

• Stats: we should support approximate stats. In large rasters to compute exact stats is a huge overhead, and probably you can work fine with approximate stats when the purpose is visualization. The approach in this PR is computing stats block by block, which is memory safe, but it could be slow for big rasters.

Do you recommend that we need to do it in this PR, or is this a suggestion for future work?

It's a suggestion to improve the script. We can add it to the backlog

[cayetanobv]

A comment about overviews strategy following GDAL defaults